key: cord-0477119-i5wgeult
authors: Hu, Catherine; Perdriau, Christopher; Mendez, Christopher; Gao, Caroline; Fallatah, Abrar; Burnett, Margaret
title: Toward a Socioeconomic-Aware HCI: Five Facets
date: 2021-08-30
journal: nan
DOI: nan
sha: 81bedb56427793856deb5b4635ac9f5268f95d0b
doc_id: 477119
cord_uid: i5wgeult

Although inequities and biases relating to people in low socioeconomic situations are starting to capture widespread attention in the popular press, little attention has been given to how such inequities and biases might pervade technology user experiences. Without understanding such inequities in user experiences, technology designers can unwittingly introduce inequities tied to users' socioeconomic status (SES). To enable the HCI community to address this problem, in this paper, we consider a wide body of research that contributes to how a user's socioeconomic status potentially shapes their interactions and user experiences. We organize that research into 20 aspects of socioeconomic status; then derive a core set of five SES"facets"(attribute types and value ranges) with differing impacts on user experiences for different SES strata; and finally, present actionable paths forward for HCI researchers and practitioners to draw upon these facets, to bring socioeconomic awareness to HCI.

In a landmark paper in CHI 2010, Bardzell introduced the concept of Feminist HCI [19] . That paper offered to the HCI community a set of foundations relevant to the user experiences of a previously under-considered group of computer users-women. These foundations provided key evidence-and theory-based scaffolding for much of the gender-HCI research that followed, ultimately greatly expanding the HCI community's awareness of gender diversity's implications for HCI.

Our goal is similar to Bardzell's, but for a different aspect of diversity that has been underconsidered by HCI-socioeconomic status (SES). As with Bardzell, this paper is not a literature survey, because we do not attempt to fully describe the research in this area. It also is not a systematic literature review or a systematic mapping paper, because we do not restrict our exploration of the literature to using only systematic techniques and do not quantitatively summarize the area. This paper is instead a synthesis of foundational literature's implications-in forms directly actionable by HCI researchers and practitioners-for bringing socioeconomic awareness to HCI.

HCI research and practice need socioeconomic-awareness foundations for three reasons. First, low-SES individuals make up a sizeable fraction of populations that technology purports to serve; for example, in 2017 in the U.S. alone, 39 million people were living below the poverty line [61] . Second, because socioeconomic status goes beyond income level-defining people's educational, occupational, and social opportunities and privileges-it can have profound effects on individuals' interactions with technology, and on technology's impacts on individuals, as we show in later sections. Third, although important prior work has empirically investigated selected low-SES subpopulations' technology needs and successes, each of these works is a unique point in the research space, but that space lacks the crosscutting foundations that would enable the HCI community to connect these unique points.

However, in 2019 the new meta-method known as InclusiveMag [112] emerged, and it provides an avenue by which to address gaps like these. InclusiveMag is a meta-method to generate a systematic, foundationsbacked inclusive design method for a diversity dimension of a researcher's choosing. InclusiveMag's bestknown "child" is the GenderMag method [29] , now in use by technologists in 40 countries (c.f. [30, 74] , and http://gendermag.org). InclusiveMag has also been used to generate inclusive design methods for several smaller academic projects, including Autism spectrum inclusivity, vision inclusivity, and age inclusivity [112] . In this paper, we harness the InclusiveMag methodology's Step 1 (Figure 1 ) to produce foundations for a socioeconomically-aware HCI, via five SES diversity "facets" (attribute types and value ranges) and actionable paths forward for HCI researchers and practitioners. [112] is a meta-method that generates specific inclusive design methods. (Note: we have changed the upper left small box from "Select software type" as per [112] to "Select technology type" to reflect the generality of the method.) This paper follows the InclusiveMag process's Step 1 (circled) to produce foundations for socioeconomic-aware HCI in the form of five SES facets.

Toward this end, this paper uses InclusiveMag's Step 1 as follows. First, we select the diversity dimension and technology type (Section 3). Second, we draw from a large collection of foundational work across multiple fields pertinent to the diversity dimension and technology type, primarily psychology, sociology, education, and HCI, to produce a set of SES facet candidates (in Section 3 and Appendix A). Third, we define criteria by which to select a final set of five facets from the facet candidates (Section 4). Finally, we provide actionable steps forward and open research questions for HCI practitioners and researchers to use in their own work (Section 5 and Section 6).

Thus, the contributions of this paper are:

• Foundations for a socioeconomically-aware HCI: A synthesis of over 200 research papers from multiple fields that can contribute to socioeconomically-aware HCI;

• SES Facets: A derivation of five facets from these foundations, that together can support actionable steps forward;

Inclusivity researchers Software practitioners

•

A many-to-many mapping between the SES facets and design recommendations from the literature;

• Actionable steps forward for researchers and practitioners for bringing socioeconomic awareness to HCI; and

• Open Questions for researchers to increase socioeconomic awareness in HCI. Researcher Self-Disclosure: Our research team is comprised of individuals who are currently mostly middle-to upper-SES individuals, but some of us have spent substantial portions of our lives as low-SES individuals. Our team has multiple races (Asian, Black, Latinx, White), with national/ethnic backgrounds from African, Asian, and North American nations. We recognize that, as academic researchers, we are in positions of privilege. We are committed to using that privilege to improve technology's inclusivity and equity for users who are disproportionately disadvantaged by today's user experience designs.

The best-studied offspring of InclusiveMag is GenderMag [29] , and as such, GenderMag paints a picture analogous to the target at which this paper aims. GenderMag's diversity dimension is gender, and GenderMag's intended use is to systematically design/evaluate problem-solving software's inclusivity across a gender spectrum.

As with any offspring of InclusiveMag, the GenderMag method revolves around a small number of facets, each backed by foundational research [29, 158] . These facets are useful for several purposes, one of which is as the basis of personas, which bring the facets to life. The five GenderMag facets capture diversity of five cognitive styles (Table 1 ) across genders. Each has a range of possible values. A facet's values at one end of the range are statistically more common among women than other people and are assigned to a persona called "Abi". The facet's values at the other end of the range are statistically more common among men than other people and are assigned to a persona called "Tim". A third set of distinguished facet values are assigned to a persona named "Pat", who has a mix of Abi's/Tim's facets plus a few unique facets. These personas are then interwoven into a systematic analytic process-for GenderMag, that process is a specialized cognitive walkthrough [180] .

Most of this paper focuses on producing a facet set analogous to the facet set in Table 1 , but for SES inclusivity rather than gender inclusivity. Thus, one end of each facet's value range will need to be more common among low-SES individuals than among higher-SES individuals, with the other end of that facet's value range more common among higher-SES individuals than among low-SES individuals. Note that this paper does not produce an entire systematic method like GenderMag. A full-fledged new method would require following both Step 1 and Step 2 of the InclusiveMag meta-method. This paper follows Step 1 only.

In the methodology prescribed by InclusiveMag, the Scope step has three parts: selecting the diversity dimension, selecting a technology type, and selecting the facets (circled portion of Figure 1 ).

First, for the diversity dimension ( Figure 1 's bottom left box) for this work, our goal is to support socioeconomically-aware HCI, so we selected "socioeconomic status". Second, for the technology type, we ultimately selected "problem-solving software", but making this selection was an iterative process. We began with a type we informally referred to as "information-spewing software", i.e., software whose main use-case is to provide information some user needs. However, as we built our facet candidates, we began to realize that the facet candidates would impact more than just an individual's attempts to find and process information. Given this realization, we broadened the scope's technology type to problem-solving software-i.e., whatever software an individual is using when they are trying to figure something out. Examples are software someone is using because they want to figure something out (e.g., spreadsheets to figure out a budget, debugging tools to track down a bug), and software that raises problems someone must solve in order to make progress (e.g., a website whose organization raises difficulties in finding information, or a software product with an obtuse user interface). Thus, the final facet set we ultimately selected both depended on and influenced the technology type.

The rest of the Scope step is the focus of the remainder of this paper: the research and analysis to derive a tractable set of SES facets with solid evidence behind them.

In the Introduction section, we defined facets as attributes with associated value ranges. Specifically for SES, we now define an SES facet as a type of attribute that has ties or correlations with an individual's SES, and its facet value as the particular value some individual has for that facet. For example, if an SES facet were "highest level of education achieved", an individual's facet value might be "some high-school". A central design goal was for the SES facets to have values near one end of an SES facet's range that are more common for low-SES individuals than for high-SES individuals (e.g., some high-school), and values near the other end of that facet's range more common among high-SES individuals than low-SES individual (e.g., graduate degree).

We also wanted each SES facet to use a wide range of values, for two reasons. First, a wide range can describe a broad spectrum of individual attributes. Second, if the facets encourage a designer to create designs that simultaneously support both ends of a facet range, then the wider the range of values, the more facet values within that range that will be supported.

To find candidate SES facets, we surveyed literature from a wide swath of topics and disciplines. We began by surveying scholarly papers (e.g., journals, papers, books, and theses) related to the impact of socioeconomic status on people's technology experiences. We started with search strings such as "Socioeconomic," "Inclusive design," and "Socio-economic Technology Design" in Google Scholar, and then expanded by following reference trails from there. Along the way, we also got feedback from experts in aspects of SES, and added literature to our survey as per their recommendations. Finally, we added statistical and census data from government and other public sources. The papers we surveyed ranged from foundational sociological research to empirical As a result, we identified 20 facet candidates. Our inclusion criteria for a facet candidate were:

• At least one scholarly paper provided evidence of the facet's potential importance to people in at least one SES stratum; and

• The facet candidate had a wide range of possible values tied to different SES strata; and

• The facet was potentially relevant to the way people in different SES strata might use technology.

The 20 facet candidates meeting these criteria are detailed in Appendix A and summarized in Figure 2 .

From the facet candidates of Appendix A, we selected a set of final facets and facet values, according to the following inclusion criteria for each individual facet:

• Criterion 1: The facet candidate is one of, or a combination of, the facet candidates from Appendix A;

• Criterion 2: At least five independent studies or papers have evidence of the facet's ties to SES characteristics;

• Criterion 3: The facet is understandable and usable by software practitioners with no background in social sciences, such as sociology, psychology, race/ethnic studies, etc.;

• Criterion 4: The facet has implications for technology that are clearly applicable to technology design decisions that could be made.

In addition, we required the final set of facets to satisfy the following criteria as a set:

• Criterion 5: A small enough set so that they could be kept in mind all at once; • Criterion 6: Together, the set can capture a large enough subset of the diverse characteristics of the population, that designing for the entire range of values for each facet can make an impactful improvement in the inclusivity of the technology.

Many of Appendix A's facet candidates arguably satisfy the individual criteria (Criteria 1-4). However, we could not select all of these facets, because of the set Criteria 5 and 6-the facet set has to be both "small enough" to be kept in mind and "large enough" to make an impactful difference.

Neither "small enough" nor "large enough" is precisely measurable, but we can inform our choices with field data from GenderMag [29] . Some GenderMag field reports suggest that keeping even five facets in mind at once may be more than some groups of GenderMag evaluators are comfortable with. For example, in two field studies of multiple teams of software practitioners [28, 69] , most teams tended to have a favorite 2-3 facets that they used for most of their evaluation, with much less attention paid to the other facets. However, the different teams did not choose the same particular facets to emphasize. For example, some have focused primarily on the risk and learning style facets [28] , others have emphasized the information processing facet [28] , and others have emphasized the self-efficacy and motivations facets [122] . Thus, this set of five facets seems to be viably small while also giving different teams the ability to focus mostly on facets that make the most sense to them.

That facet set has also turned out to be large enough to be powerful [28, 122, 175] , with a very low rate of false positives [29, 175] . This suggests that a facet set with about five elements satisfies Criteria 5-6, so we aimed for about five facets for our SESMag facet set. Example potential impact: Users who do not comprehend implicit assumptions, cultural references, vocabulary, or complex structures in the system's communication language (e.g., English) will be less successful in using the system than users with high fluency.

Low-SES example: Low-SES people, especially migrants and ESL (English-assecond-language) speakers, are limited to literary engagement mainly with individual words or clauses (menus, receipts, etc.), and have low literary engagement/experience with documents such as newspapers [47, 101] .

Users' belief in their own ability to exert influence over technology's positive or negative outcomes and interactions. Can relate to their attitude toward authority figures and the technology's outputs as an authority [32, 33, 46, 60, 76, 77, 81, 86, 91, 92, 109, 130, 134, 136, 137, 163, 164, 166, 181] . Facet's value range: Low to high perceived control.

Example potential impact: Belief in the Just World Hypothesis implies that authority systems are legitimate, which will lead to a willingness to see technology barriers as negative outcomes not to be questioned Low-SES example: Unquestioning attitude toward authority and perception of low control can prevent negotiation of parameters. For example, low-SES students more likely to turn in incomplete work instead of negotiating for an extension [33, 77, 130] . Attitudes toward Technology Risks, Privacy, and Security: Willingness to take technological risks, including all kinds of technology-based risks, ranging from using "mysterious" unfamiliar software features, to risking privacy loss, risking device security, etc. [9, 16, 25, 67, 68, 107, 110, 145, 159, 167, 174] . Facet's value range: Low to high tolerance for technological risks.

Example potential impact: Users fearing privacy risks (revealing identity or location) will be unlikely to use technology.

Low-SES example: Undocumented Latinx immigrants are at risk of deportation/arrest if privacy is breached online [68] .

We ultimately chose exactly five, namely the five facet candidates that best satisfy the individual criteria and/or best contribute to group Criteria 5-6. Table 2 overviews the final five facets, and the subsections that follow present each of these final facets, their importance to problem-solving technology's support of SESdiverse populations, and the extent to which each facet satisfies the above criteria.

From the candidate facets in Appendix A, we selected Technology Self-Efficacy as one of the final facets.

Self-efficacy is the belief an individual has about their ability to perform an upcoming task in order to achieve a goal [18] . Self efficacy can be applied to many contexts, and here we use technology self efficacy, an individual's belief in their own abilities to interact with unfamiliar technology. Self-efficacy can have numerous effects on the individual's ultimate success with the task, including whether they blame themselves for difficulties they encounter, and their willingness to persevere in the face of difficulty and try different approaches to the problem if their first attempt fails [18] . Such effects have been shown multiple times with technology/computer self-efficacy (e.g., [158] ). Many studies have established differences in technology/computer self-efficacy tied to people's SES level.

Sometimes this phenomenon is also tied to minority ethnicities or races. For example, Black Americans are disproportionally more likely to experience poverty than White Americans [136] and in a study of computer selfefficacy in Historically Black College and University (HBCU) students vs. Predominantly White University (PWI) students, HBCU undergraduate computer science students' computer self-efficacy was significantly lower than PWI students [99] . In addition, African American students overall had significantly lower computer self-efficacy than White students, without regard to which institution they attended [99] . However, Black Americans are just one example of this phenomenon: Hatlevik et al. found that students' technology self-efficacy was significantly influenced by their SES and their prior experiences with technology, in a study conducted across 15 countries [72] . Experience affects self-efficacy [18] , and some have found that lack of access to modern technology (discussed further in Section 4.2) is part of the reason for low-SES populations' lower technology self-efficacy. For example, one study found that students from low-SES families expressed lower confidence in their ability to use Information and Communications Technologies (ICTs), often due to a lack of internet and/or computer access [171] . Low-SES populations often must share devices or rely on public devices [188] , and in one study of 345 students, 8 of the 11 students who did not use computers outside of school came from low-SES families [171] . In addition, when growing up, low-SES students do not have as reliable access to in-school technologybased learning opportunities, bringing them fewer opportunities to experience technology [104, 106, 178] . These reasons may contribute to low-SES populations' more negative computer beliefs compared to their peers from middle and high-SES populations [72, 172] .

As the previous paragraph suggests, the value of one facet can exacerbate another facet's effect, even when the two facets' definitions are independent. In the example above, although the facet of technology self-efficacy is defined orthogonally to the facet of technology access, lack of access can affect an individual's feelings of technology self-efficacy through the mediating factor the two have in common, namely experience. Another example is the facet of perceived control, with the factor in common being prior successes. As Bandura's writings point out (e.g. [18] ), an individual's history of successes helps to determine their self-efficacy with similar future endeavors. Likewise, an individual's past successes at controlling technology outcomes affects their perceived control over future technology outcomes, as we discuss in Section 4.4. These facets individually or in combination may leave low-SES individuals with a sense of inevitable failure in their computing tasks.

As Huang et al. summarized, the interplay between an individual's technology self-efficacy and their SES level is not surprising, given the social science underpinnings in common. As Huang et al. put it, "Four … factors which may affect <self-efficacy> include attainment, experience, social persuasion, and physiological factors (Bandura, 1995) . Interestingly, these four factors may all be affected by family SES as well" [75] .

These findings suggest (at least) the following research hypotheses for a socioeconomic-aware HCI: H1: Low-SES individuals' low technology self-efficacy may lead them to expect that failing is inevitable when attempting technology tasks that seem complex. H2: Low-SES individuals will be more likely than higher-SES individuals to blame themselves when they run into technology errors or other negative outcomes with technology. H3: To support low-SES individuals in using unfamiliar technology, it will be important for problemsolving software to clearly state where/how to get help if needed.

H4: Low-SES individuals will be more likely to actually fail tasks with problem-solving software than high-SES individuals, because their low technology self efficacy may act as a "self-fulfilling prophecy". Regarding the facet inclusion criteria from the beginning of Section 4, Appendix A enumerates how selfefficacy satisfies Criteria 1, 2, and 4. Regarding Criterion 3 (understandability), self-efficacy's usage as a Gen-derMag facet 1 provides evidence of its understandability in the literature documenting numerous software practitioners successfully using GenderMag's self-efficacy facet (e.g., [28, 122, 175] ). Technology self-efficacy's usage with GenderMag also provides encouraging evidence that this facet may contribute strongly to the set Criterion 6 (impact). As a facet in the GenderMag set, self-efficacy has been impactful in identifying genderinclusivity issues in software. For example, in one study of Open-Source Software's (OSS) inclusivity issues, the self-efficacy facet helped to identify 88% of the inclusivity issues that the OSS practitioners found [122] . In another study of fixing inclusivity errors in a search engine's interface, self-efficacy inspired the design changes behind half of the fixes [175] .

The second facet we selected from the candidate facets is Access to Reliable Technology. This facet includes both (1) access to technological devices and the internet, and (2) reliability of that level of access. We selected this candidate as a facet for two reasons: (1) an individual's SES affects the technologies available to them, especially hardware devices and internet; and (2) the kind and the reliability of hardware and internet an individual can use are prerequisites to the use-cases in which an individual can participate, and the quality of their user experiences in doing so.

First, we consider technology infrastructure. The technology infrastructure available to an individual, such as hardware devices and internet access, varies for different SES strata [58, 68, 104, 153, 170, 178, 188] . For example, some low-SES families cannot afford to purchase a personal device for every member, and some cannot afford to pay a monthly internet fee [139, 177, 188] . Even if a low-SES individual owns their own device, they may be able to use the device only in public spaces like the library, such as with families who own laptops but lack home broadband [188] . The United States Census Bureau has reported that, although 62% of low-SES households have computers, less than half have internet access, and many others have poor connectivity due to slow service, outdated technology, or lack of payment [58, 139] .

These infrastructure differences across SES strata can impact how individuals try to use technology, what they try to accomplish with it, and what their user experiences are. For example, because of their limited access to technology, many low-SES individuals use devices in more restricted ways than higher-SES individuals do. In one higher education study, most low-SES students used the internet primarily for communication through email, discussion forums, and downloading music [153] . In contrast, high-SES students used the internet for a far more comprehensive range of activities, including reading online books and newspapers, playing video games, and seeking information on goods or services [153] . Several other works have reported that low-SES students are less likely to have access to computer-related courses in their schools [104, 106, 178] . Further, some low-SES populations rely mainly on mobile phones to access the internet (e.g., [68] ), which restricts the user experiences available to them.

These phenomena suggest the following hypotheses about low-vs. higher-SES individuals' technology usage:

H5: Low-SES users will not be as familiar with new technologies and features as higher-SES users, because new technologies are expensive and low-SES users therefore have less access to them.

H6: Technology that supports sharing will be especially useful to and favored by low-SES users. H7: Software and websites that provide good user experiences on mobile phones will be especially useful to and favored by low-SES users. Appendix A enumerates how this facet satisfies the inclusion criteria for Criterion 1 (was a facet candidate), Criterion 2 (at least 5 different studies/papers), and Criterion 4 (has implications for technology design). We expect it to also satisfy Criterion 3 (understandability even without social science background), because occasional bouts of unreliability and loss of internet access occur regularly in technology fields, and software practitioners are likely to have experienced these issues themselves. However, factual evidence of the expected understandability will require field-testing.

Appendix A include mastery of language, culture, and education as separate facet candidates. Because the literature shows numerous interconnections among these three facet candidates [47, 88, 118, 131, 150] , we combined them into one facet to capture an individual's fluency imparting and understanding the language the technology uses.

Low language literacy can disadvantage users when navigating technology [39, 70] , and low-SES users are more likely than higher-SES users to have low language literacy [47, 53, 101, 121] . For example, research from both the US and Europe shows that the lower an individual's SES, the lower their language literacy is likely to be, even in their native language (e.g., [2, 53, 88, 121] ). One factor is less literacy engagement. For example, some low-SES individuals engage with reading at only the level of individual words and clauses, such as with menus, receipts, or texts-but not at the level of whole sentences or paragraphs such with newspapers or novels (e.g., [101] ).

As the literary engagement levels help to clarify, mastery of a language goes beyond grammar and standard vocabulary. It also can include the idioms, specialized vocabulary, cultural references, and sentence structuring that some particular technology requires for communicating with it. Sometimes the technology's communications are in English, sometimes in other languages, and sometimes in specialized languages involving specialist vocabularies, analogies or references to an assumed shared culture. As an example, Stack Overflow (stackoverflow.com) uses "techie English"-it uses tech-specialist vocabulary ("parser"), assumes shared tech-specialist experiences that obviate the need for details ("look at the variables at that point"), and requires college-level reading ability 2 .

Education is a critical factor in opportunity for literacy-learning activities. More educated individuals have higher literacy than less educated individuals, across all age groups and across over 40 countries [23, 40] . However, low-SES individuals generally receive lower quality education than higher-SES individuals [77] . Further, low-SES individuals tend to receive lower technology-related education than higher-SES individuals [43] . For example, one study showed a multitude of significant relationships between a child's family-SES and their technology-related achievements [1] . Some of these relate directly to language literacy, such as the differences by SES level in processing and combining textual information from information products [1] .

The above pertained to education's impacts on technology/digital literacy. There are also issues on education with technology, and some pertain to culture.

An individual's education experience is mediated by the culture with which they identify, and many low-SES individuals are members of minority cultures. Research into supporting education and literacy with technology has shown several ways low-SES minorities can gain literacy through technology without loss of ties to their own cultures. For example, some low-SES individuals in minority cultures benefited from a multimodal approach to literacy learning with different modalities (audio, visual, digital, body movement, etc.) through which people can acquire language [118] . In another example, a do-it-yourself approach, some mothers of children experiencing life on both sides of the U.S.-Mexico border used multiple digital tools to help educate their children in English fluency while still staying connected to Mexican culture. For example, one mother had her child watch Spanish science videos, listen to English music, and play Xbox games in English, for the mother to monitor the child's bilingual and bicultural exposure and comfort level [118] . Another study showed that Massively Multiplayer Online Games (MMOGs) can serve as an effective tool in helping learners of English-as-a-second-language pick up American culture in an immersive, culturally safe setting [150] . In that study, Shahrokni et al. performed a narrative inquiry into how native and non-native English speakers interacted in the game Stronghold Kingdoms, finding that through teamwork, shared goals, and socializing, players were able to gain fluency with the online community's linguistic and cultural norms [150] . These findings about communication differences across SES levels, and their interdependencies with language literacy, culture, and education, suggest the following hypotheses:

H8: When technology uses complex language structures, advanced vocabulary, or cultural references, low-SES users will be more likely than higher-SES users to experience problems with the technology.

H9: Technology that is culturally flexible (e.g., can be made to reflect the culture with which an individual most identifies) will be more inclusive across SES levels than technology that is not culturally flexible.

Regarding the inclusion criteria, this facet satisfies Criteria 1, 2, and 4 for the final facet set, as detailed in Appendix A. We expect the language and education aspects to be understandable and directly usable (Criterion 3), but the culture aspects' relationship to communication literacy may require more social science background than some software practitioners have. Empirical research of its usage in the field will be required to answer this question.

The fourth facet is a combination of two of the facet candidates, Perceived Control and Attitude toward Authority. Although each of these facet candidates varies with an individual's SES and each has potential impacts on how an individual interacts with technology, we chose to combine them into one facet in that each of them interrelates with the other in ways that are difficult to separate.

Perceived Control refers to an individual's belief that they can exert influence over future events [123] . Research has shown that low-SES populations often feel a lack of agency or control over their lives [130, 136] . An individual's perception of control over their lives interacts with their attitudes toward and behaviors with authority figures, and these too vary by SES. For example, many low-SES individuals are not in positions of power, which not only decreases their perceptions of control, but also comes with a requirement to comply with the dictates of authority figures [60, 71, 166] . Consistent with this reality, decades of research have strongly correlated low-SES individuals' perceived lack of control with a tendency to be accommodating to authority figures (e.g., [164, 166, 186] ). Further, low-SES individuals lack experiences, practice, and the cultural capital to interact with authority figures as their equals and are also less likely than higher-SES individuals to be overtly critical of authority figures [163] .

An individual's perception of control over their outcomes and their attitude toward authority have direct ties to their behaviors with and around technologies. For example, when low-SES students run into technology issues, they are less likely to request help or extensions from teachers because they believe that the teacher will not act upon their requests-instead they simply turn in their work as far as they got [130] . In contrast, higher-SES students tend to believe that their teacher will accommodate their technological issues and ask for extensions, enabling them to finish their work [130] . Other studies have reported similar phenomena [33, 77] .

In addition to the dependence of some low-SES individuals' technology usage on what authority figures allow, some people regard computers as "authorities" themselves or are willing to grant computers authority over them in some circumstances [21, 78, 94, 152] . This attitude, when coupled with an attitude of deference to authority, suggests that low-SES computer users might be more accepting of negative computer outcomes (e.g., web pages with greyed out "submit" buttons, paths leading to error messages, etc.) than people who are more critical/questioning of authority.

Some readers may wonder why, when the computer erects a roadblock (e.g., a dead-end in a navigation path or an inexplicable error message), low-SES users would be less likely than higher-SES users to "push back" and try to find another route to reach a solution. Besides the relationships to control and authority already discussed, another answer may lie in the Just World Hypothesis.

In order to cope with feelings of lack of control and powerlessness, low-SES individuals are also more likely to hold system-justifying beliefs: beliefs that validate, internalize, and legitimize existing hierarchies. The Just World Hypothesis (JWH) is a type of system-justifying belief, in which good things happen to good people and bad things happen to bad people [8] . Although it may seem paradoxical, an individual with low SES is more likely than someone with higher SES to have a high belief in the JWH [76] . For low-SES individuals, among the negative costs associated with system-justifying beliefs are internalization of inequality and greater self-blame for societal disadvantage [109, 181] . However, having system-justifying beliefs like the JWH is positively correlated with an individual's perception that they have control over their future and that advancement is possible. This greater perception of control can mediate mental and physical health benefits, such as having a boosted self-esteem, for low-SES people [109, 181] . Thus, low-SES individuals receive psychological benefits from perceiving society's distribution of outcomes as fair-even if that distribution puts themselves at the bottom [109] . Although multiple theories consider why low-SES people hold system-justifying beliefs, most theories agree that system-justifying beliefs serve as a response to an unfair social system, i.e., low perceived control [60, 80, 186] .

The combination of perceiving little control over outcomes in life, deference to authority, and system-justifying beliefs can interact with technology usage in nuanced ways. For example, these attitudes can affect people's "planfulness"-willingness to take actions now that are expected to bring benefit in a more distant future. Thus, low-SES individuals are less likely than high-SES individuals to plan ahead [22, 48, 111] . For example, Laurin et al. found that low-SES individuals were more likely to work toward future goals when they believed that awards would be distributed fairly, i.e., they would get what they deserve. (On the other hand, higher-SES individuals worked less while having the same belief that awards would be distributed fairly [92] .) Even when society is not fair to an individual, low-SES individuals who believe that it is fair have been shown to be more willing to invest time working toward long-term goals (e.g., Figure 3 ) than those who do not believe it is fair [92] . These perceptions can interact with the problem-solving software that is our scope, because much of it involves planning ahead. For example, complex software features require users to invest time in learning new features that will (hopefully) pay off later. From these foundations on how perceived control and attitude towards authority differ between low and high-SES individuals, we derive the following hypotheses: H10: When low-SES users run into errors, dead-ends, or unfavorable outcomes, they will be more likely than higher-SES users to accept them as outcomes to which they have no recourse.

H11: Low-SES users will be less likely than higher-SES users to engage in the kinds of planfulness or investing in the future needed to learn complex technological feature sets, or to plan a complex strategy for overcoming a problem with technology.

Appendix A enumerates how this facet satisfies Criteria 1, 2, and 4 in the final facet set's inclusion criteria. We also expect the facet to satisfy Criterion 3, because many people have experienced the frustrations of lack of control over some aspects of technology, but field investigation is still needed to gather empirical evidence of this criterion being satisfied.

Several of the candidate facets in Appendix A pertained to elements of risk that can arise in using technology.

Two of these technology risks-privacy and security-are so important, an entire subfield (or two subfields by some counts) investigate(s) them. However, privacy and security are not the only types of technology risk that are particularly salient to low-SES individuals. Thus, we collected elements pertaining to risk from all the candidate facets in Appendix A, including privacy and security. We named the facet "Risk, Privacy, and Security"

to keep explicit that Privacy and Security are not the only types of risks that should be considered in bringing socioeconomic awareness into HCI. Starting with privacy and security risks, some fears common across all socioeconomic strata relating to privacy/security are the risk of identity theft, of online financial fraud, and of hackers who might steal or take over information such as passwords [68, 159] . For example, in a recent Pew Research study [16] , 28% of Americans say they have suffered from someone putting fraudulent charges on their credit or debit card, someone taking over their social media or email accounts, or someone trying to open a credit line or get a loan using their name.

However, some demographic groups that are disproportionately in low-SES strata worry even more about such risks, and for good reason-such risks materialize for them more often. For example, Black adults have had someone take over the social media or email account at almost three times the rate of White adults and are also more likely than White adults to have had someone attempt to apply for credit or a loan using their name [16] . Black Americans are also more likely than White Americans to believe the government is tracking them online, and Black and Hispanic Americans are more likely than White Americans to be concerned about what law enforcement officials, employers, and family and friends know about them [16] . In a study of undocumented immigrants, many participants were hesitant to post images of themselves on Facebook, for fear that they would be recognized and then arrested or deported [68] .

Part of the reason for this heightened caution has been the higher prevalence of police and government surveillance in low-SES communities. In some low-SES communities, surveillance is constant and brings disproportionately heavy consequences for people in those communities (e.g., greater rates of arrest and policeor government-related deaths for Black and Hispanic people, who are disproportionately low-SES). Low-SES individuals can also feel particularly susceptible in other ways that arise less often than with higher-SES individuals' privacy and security concerns. One group of such issues arises because of the prevalence of shared devices among low-SES individuals (e.g., [188] ), as mentioned in Section 4.2. In these situations, individuals may not trust the people who share their devices to not view their browsing histories, data, apps, etc. [145] .

To guard against risks like these, low-SES individuals use a variety of practices. In addition to common security/privacy practices like phone locks and passwords, low-SES individuals also sometimes simply avoid engaging with technology [25, 145] . Another tactic low-SES individual use is to delete extensively. In one study, participants safeguarded privacy by removing sensitive content. For example, some deleted entire threads or histories and others deleted specific chats, media, or queries [145] . These strategies helped them maintain privacy while simultaneously adhering to cultural expectations such as sharing their mobile phones with their social relations (e.g., [145] ). Another study of 28 low-SES New York youth likewise found that its participants were very cautious with their online information [107] . Among their privacy measures were asking friends to remove social media tags identifying them, because they feared consequences such as family drama or compromised employment [107] .

Beyond privacy and security, low-SES individuals experience other technology-related risks at a disproportionately high rate. For example, a risk particularly prevalent for low-SES individuals is that of unreliabilitysince low-SES individuals tend to use older, less reliable devices, and less stable internet connectivity, the risk here is that technology will fail them before they can succeed at what they are trying to do with it [174, 188] . Low-SES individuals' risk of unreliable internet and devices is intertwined with their reduced access to technology (Section 4.2).

Even when the technology holds up, low-SES individuals risk experiencing disappointing outcomes from using technology. For example, online job searches tend to support professional and highly qualified job seekers, but not low-SES individuals [52] . Other outcomes have been reported to be especially prevalent since the 2016 U.S. election, due to an increasingly xenophobic online atmosphere fueled by political tensions. As such, many low-SES participants reported not engaging in public online discussions, for fear of doxxing or online harassment [68] .

Another risk issue more prevalent among low-SES individuals than higher-SES individuals comes together with cultural values. For example, in one case study of low-SES households in three South Asian countries [145] , privacy was sometimes regarded as a malicious way of hiding information from other people, which ran counter to their cultural values. As such, it was considered disrespectful for a woman to refuse to hand over her phone to men or elders. Cultural issues also arise with low-SES individuals who are also members of racial or ethnic minority groups, where risks of technology incompatibilities with the cultural and social norms within their group are different from norms of the society that created most of the technology they use [9] .

Finally, alone or in combination with many of the above risks is the risk of wasting time without gaining benefit. Low-SES families sometimes must make ends meet by working multiple jobs [7, 12, 38] . They also tend to have multiple, competing demands on their time, such as requiring extra time to acquire basic necessities for their children. For example, their time often goes to complicated logistics due to use of public transportation and time-consuming dealings with bureaucracies for income, food stamps, etc. [189] . Thus, when they perceive risk of failure due to one of the risks above, they also perceive a risk of wasting their limited time on work that will not succeed.

These findings suggest the following hypotheses about technology usage differences among low-SES vs. higher-SES individuals:

H12: Low-SES individuals will be less likely than high-SES individuals to successfully complete tasks with problem-solving software due to aversion to technological risks (e.g., aversion to potentially wasting time on new/unfamiliar features, aversion to potential loss of privacy, …). H13: Low-SES individuals will be more likely than high-SES individuals to perceive technology features as being risky, and therefore will be less likely to use those features. Appendix A enumerates how the Risk/Privacy/Security facet satisfies Criteria 1, 2, and 4 in the final facet set's inclusion criteria. As the Appendix shows, even without considering the privacy and security aspects, the risk aspect satisfies the final facet set inclusion criteria. As with the self-efficacy facet, insights into this facet's ability to fulfil Criteria 3 and 6 can be gleaned in risk's presence as a GenderMag facet [29] . In the GenderMag context, the Risk facet has been impactful (Criterion 6), which suggests that it has been understandable enough for software practitioners to use (Criterion 3). For example, in one study of Open-Source Software's inclusivity errors, software practitioners used the risk facet to help identify 71% of the inclusivity errors they found [122] . In another study of fixing inclusivity errors in a search engine's interface, software practitioners used the risk facet as a foundation behind all of the inclusivity errors fixes, they devised [175] .

We have hypothesized 13 phenomena that these facets can reveal about SES-diverse users' experiences with technology, especially with problem-solving software. But when a facet reveals such a phenomenon, what then?

The literature offers the beginnings of answers to this question, in the form of recommendations on how to design technology for some low-SES subpopulations in specific situations. Below, we present the design recommendations we located, ordered by the five facets, and Figure 4 presents a summary of these recommendations.

The literature contributes two design recommendations relevant to supporting SES-diverse users' Technology Self-Efficacy facet values. One relates to Hypothesis H1. Recall that H1 is that low-SES individuals' low technology self-efficacy may lead them to expect that failing is inevitable when attempting technology tasks that seem complex. As a result of an investigation into how parents in economically depressed communities find learning resources for their children, Roshan et al. recommended that learning resource search technologies should include search scaffolding, because low-SES parents were not using the same search vocabulary as the technology expected [141] . We hypothesize that support for such scaffolding will help support low-SES individuals' low technology self-efficacy, because the design recommendation may prevent users from failing to locate the learning resources they seek.

The other design recommendation is a collection of recommendations about self-efficacy-related design remedies that were revealed for GenderMag's self-efficacy facet. The GenderMag project has an emerging design catalog that enumerates real-world design remedies by facet, including real-world usage and/or evidence behind each design remedy [66] . These remedies range from labeling potential tasks with difficulty levels to adding concreteness to adding process scaffolding. These self-efficacy-related design remedies for supporting gender-diverse populations are also potentially relevant to supporting SES-diverse populations.

The literature offers six design recommendations for SES-oriented situations pertinent to the Access to Reliable Technology facet. The first is a collection of recommendations specific to supporting homeless individuals, for whom access to reliable technology is often very problematic [183] . This collection recommends that technology should be designed to have low cost; should be outdoor resilient (e.g., against wet and cold); should have access to flexible non-infrastructural energy sources (e.g., solar); and should have 24/7 reliability for emergency use [183] . In addition, Le Dantec and Edwards' research revealed that cell phones were the preferred device for many homeless individuals-both as a communication device and a symbol of belonging to society [93] . Therefore, they recommended that software applications be designed to work on cell phones. More generally, Wyche and Murphy's recommendation expanded on reducing costs, and recommended standardizing mobile phone designs. They argue that this would result in more durable and cost-effective phones, which can in turn increase accessibility for low SES populations [185] .

Some design recommendations, rather than focusing on increasing access to or reliability of a technology, focus on providing alternatives to a technology that is failing a low-SES individual. For example, Vashistha et al.'s research into low-SES blind people recommended designing voice-based forms of social media. This recommendation not only relates directly to the Access to Reliable Technology facets by providing an alternative Figure 4 : Solid lines denote a direct connection from the referenced paper to the facet, whereas dashed lines denote a connection from the referenced paper to our facet-derived hypotheses from Section 4.

Access to Reliable Technology Communication Literacy/ Education/ Culture

Use fewer colloquialisms in language [70] Design phone-friendly apps since they more commonly used by low SES [93, 168] Scaffold search keywords to minimize vocabulary barriers in the technology [141] Use firsthand experiences from real people when designing applications that communicate dietary recommendations [102] Provide separate communication channels protected from surveillance [183] Design technology to be socially compatible with localized needs, have content appear in the local language [79] Cost transparency, privacy transparency, use local platforms, show ratings by own friends [51] To give users more control, incorporate a back and home button for shorter and longer recovery lengths, respectively [37] Design activities that don't require culturally specific knowledge [9] In classroom settings provide visuals to offer a lower barrier to participation [116] Provide the visual affordance as a way to supplement low language literacy [37, 54] Present information in a visual manner (red light for urgent and green for success) [98] Support on-demand information hiding [68] Use plain language to minimize cognitive load [39] Provide access to iPads to increase children's literacy [26] Consider students' cultural and linguistic backgrounds when developing learning tools [131] GenderMag Design Catalog's design recommendations for selfefficacy and risk facets [66] Attitudes toward Tech Risks,

Standardize mobile phone designs to make handsets more durable and cost-effective [185] Provide voice based options to blind users to cover come the cost of screen reader software [169] Provide an offline way to share knowledge and experiences when technology becomes unreliable [103] Make technology outdoorresilient, flexible power decoupled from infrastructure [183] Provide alternative ways to access unreliable digital media [73] Support telemedicine as an alternative when transportation to clinics is not reliable [160] to expensive screen reader software, it also increased their social skills [168, 169] . In another example, Marcelino-Jesus et al. recommended providing both online and offline spaces for students to access and share knowledge, since online technology was not always reliable [103] . Along similar lines, Hebert et al. recommended making physical (offline) copies of information available [73] , to support low-SES individuals who cannot afford the cost of reliable access to WiFi/technology. Another investigation considered using cell phone technology to mitigate failures in transportation technology. Specifically, in an investigation into how low-income communities' access and receive healthcare, Tang observed that low-income communities are more likely to have access to communication technologies, like phones, than to transportation technologies, like cars or reliable public transportation. Based on this observation, Tang recommended that telemedicine be made available as an alternative to in-person doctor appointments, since not low-SES patients may not have access to transportation technologies that are reliable or that can fit their scheduling constraints [160] .

The literature revealed ten recommendations for the Communication Literacy/Education/Culture facet. The ten recommendations fell into three categories: recommendations about English language/vocabulary inclusivity, recommendations on visuals and imagery for inclusivity, and recommendations about cultural inclusivity. Since many low-SES populations have lower literacy in the language the technology is using to communicate (usually English), some recommendations relate to reducing literacy barriers by maintaining simplicity in the non-native language used in the technology. For example, Guo's study into how non-native English speakers learn computer programming revealed barriers in reading and writing code, as well as understanding technical vocabulary [70] . Note that two non-native languages arise in Guo's study: English language, and technical vocabulary. A third literacy issue also arose in Guo's study, namely English cultural references. Guo recommended using simplified English, including simple sentence structure, fewer colloquialisms, and fewer references to English culture-specific topics [70] . Cheung likewise recommended using plain language in professional and technical software interfaces, pointing out that plain language can help minimize the cognitive load that low-SES individuals may endure when using these interfaces [39] . Instead of reducing such literacy barriers, Kaur et al. [79] and Maitland [102] recommended removing them entirely, by having technology communicate in the user's local language.

Rather than reducing/removing literacy barriers, some researchers recommended approaches to help users overcome them. One example was platform-oriented. Breazeal et al. found that children who used iPads in an educational context had an increased literacy and vocabulary when compared to peers without access to iPads. Thus, they recommended providing children with iPads for better learning outcomes [26] . Another example was scaffolding-oriented. Returning to Roshan et al. (also discussed in Section 5.1), they reported that low-SES parents often could not find their desired online educational resources because of vocabulary mismatches in internet searches. Therefore, they recommended that search keywords be scaffolded with autocomplete capabilities to suggest to low-SES parents search keywords that are relevant to their search [141] .

One way to reduce literacy barriers is to replace use of written language with use of appropriate visuals and images. For example, Chaudry et al.'s study on designing a mobile health application for low-literacy and lowtechnical literacy users found that their participants favored visual elements such as non-text-based graphical user interface widgets and large radio buttons over text [37] . Similarly, Liu et al.'s qualitative study involving a prototype mobile application to support communication between low-SES parents caring for high-risk infants and doctors, recommended incorporating recognizable visualizations [98] . By communicating a red light for high alert and green light for success, researchers were able to mitigate the low-literacy and language barriers that low-SES parents faced upon communication with their pediatric providers. Nacu et al. also recommended using more visual interactions. When teaching low-SES Latino students on an online platform, they decided to use visual reactions, such as emojis instead of open-ended textual commenting. They reported that this decision improved the fit to the Latino students' cultural norms and presented a lower barrier to participation [116] .

The literature offered several recommendations centered on cultures. For example, in a study of information and communication design in rural, socioeconomically underdeveloped Indian villages [54] , Dutta and Das recommended using community-generated visual elements that are culturally appropriate. In a study of African-Aemrican students who lived in mid or low-SES neighborhhods, Rader et al. also recommended that technology take into account individuals' cultural and linguistic backgrounds [131] . Their tutoring tool was cognizant of African American communities' literacy conventions in both African-American English and in narrative storytelling. As a result, the tool helped low to mid-SES African-American children pick up "school English" literacy by enabling the children to connect to the concepts in their own dialect and culture. In an investigation of healthrelated behavioral change technologies, Maitland observed low-SES families' motivations to make positive dietary changes [102] . In order for health messages to support actionable change instead of being designed to simply motivate change, Maitland recommended that health messages be delivered in a community-based manner [102] . For example, when trying to convince parents to introduce healthy foods to a child, Maitland found that a forum with firsthand experiences is preferable to a generic website that lists potential strategies. Finally, Ames and Burrell suggested avoiding cultural mismatches by designing technology with less culturespecific knowledge, such as having a treasure hunt instead of a Harry Potter activity in MineCraft [9] , thereby removing barriers to participation based on an individual's prior readings or social activities.

We located only one design recommendation in SES-related literature pertinent to the Perceived Control/Authority facet. In this design recommendation, Chaudry et al. points out the importance to low-SES individuals of a classic design technique: that every page should incorporate back and home buttons, to enable users to recover from mistaken or dead-end interactions [37] . As many others have pointed out, this classic technique relates directly to user control, and Chaudry et al. reported it in their work with low-income users. The particular importance to low-SES users relates to Hypothesis H10: When low-SES users run into errors, dead-ends, or unfavorable outcomes, they will be more likely than higher-SES users to accept them as outcomes to which they have no recourse. The addition of controls that consistently and explicitly offer alternative pathways to a solution may help to mitigate the hypothesized H10 effect.

A number of design recommendations also addressed technology related to the Attitudes toward Risk, Privacy, and Security facet. As Guberek et al. pointed out for undocumented immigrants living in the United States, using technology for social media and communication comes with risks of fear of deportation and offline threats [68] . As a result, Guberek et al. recommended wide support for on-demand information hiding, so that sensitive information on their devices will not be released if detained. Additionally, Guberek et al. recommended making information flows between the technology and user more transparent in order to communicate potential risks upfront. Woelfer et al.'s research on designing technology for homeless people made a similar recommendation to establish privacy within technology. In order to protect homeless people from surveillance, Woelfer suggested for technologies to employ separate communication channels [183] . Dillahunt et al. researchd low-income individuals who live in transportation scare environments and use ridesharing services. They found that low-income individuals faced barriers such as low application transparency and low digital literacy. The study concluded by suggesting that in order for low-income individuals to trust these platforms, ridesharing service providers should coordinate with local businesses to design applications that build trust within a community. By installing public kiosks in intermediate locations, low-income users will better understand how their information is being collected due to integrated community engagement [51] .

In addition, because GenderMag also uses the Risk facet, the design recommendations in the GenderMag Design Catalog that are pertinent to Risk could also be applicable to this facet [66] .

The facet set we have presented can be seen as a distillation of many researchers' prior contributions into the set of vocabulary and concepts we term "facets." Our goal in doing so is to provide these cross-cutting foundations in actionable form to HCI researchers and practitioners. Actionability is intertwined with validity and utility: considering whether the facet set is valid, so that it can be acted upon in good conscience; and whether the facet set has utility, so that acting upon it is time well spent.

In formally evaluating a facet set like the one presented here, several validity and utility perspectives are possible. The first two we consider are internal validity and completeness.

The answers to these two questions are straightforward: (1) Internal validity is true by construction, and (2) completeness is obviously false (counterexamples in Appendix A). More specifically, internal validity asks whether these are facets whose values actually vary with socioeconomic status. Each facet satisfies this criterion because we selected only facets that satisfied it, by Criterion 2 (Section 4). The literature review summarized in Appendix A details the evidence for each facet. Completeness is whether the final set of five facets contains all of the important attributes of low-SES individuals. We know that it does not. For example, Appendix A suggests that there are at least 20 facets potentially useful for a socioeconomic-aware HCI, and our facet set size is only five.

Given that our facet set is a valid subset of SES-diverse characteristics, the next questions are whether this subset is usable and effective. The usability question is open. Answering the usability question will require HCI practitioners to start using it in the field. As creators of other InclusiveMag-guided facet sets have found, facet sets may need iteration to achieve practical usability. For example, in an academic study of InclusiveMag usage [112] , the teams iterated as they attempted to apply their facets to real products. Through their iterations, some teams found that some of their facets were too concrete and needed to be generalized, while others were too general and needed to be more concrete. Likewise, the GenderMag method started with five evidence-based facets, later dropped one as a result of field testing, and still later added a different one as a result of more field testing [29] .

The question of the set's effectiveness in practice also remains open. Recall Criterion 6 (Section 4): does the facet set indeed capture a "large enough" subset of SES-diverse characteristics, that designing around them makes an impactful improvement in the SES-inclusivity of the technology? Approaches that have been used to answer these questions for GenderMag facets suggest similar possible steps forward for SES facets. One example is Vorvoreanu et al.'s recent before/after study on a product redesign based on gender facets [175] , and another example is longitudinal testing of the gender facets and the artifacts/methods derived from them (e.g., [74, 122] ).

Another utility question is extensibility, and one particularly challenging extensibility question is how the SES facets come together with intersectionality. Intersectionality is a relatively recent term to HCI [148] to capture the notion of multiple identities occurring in one person. An individual in any SES group also belongs to other identity groups. For example, everyone is in a racial group(s), has a gender identity, and so on. HCI's increased attention to intersectionality is because slicing populations up by only one dimension (e.g., by gender only) Table 3 : Summary of the hypotheses derived for each facet.

Self-Efficacy H1: Low-SES individuals' low technology self-efficacy may lead them to expect that failing is inevitable when attempting technology tasks that seem complex. H2: Low-SES individuals will be more likely than higher-SES individuals to blame themselves when they run into technology errors or other negative outcomes with technology. H3: To support low-SES individuals in using unfamiliar technology, it will be important for problem-solving software to clearly state where/how to get help if needed. H4: Low-SES individuals will be more likely to actually fail tasks with problem-solving software than high-SES individuals, because their low technology self efficacy may act as a "self-fulfilling prophecy".

Access to Reliable technology H5: Low-SES users will not be as familiar with new technologies and features as higher-SES users, because new technologies are expensive and low-SES users therefore have less access to them. H6: Technology that supports sharing will be especially useful to and favored by low-SES users. H7: Software and websites that provide good user experiences on mobile phones will be especially useful to and favored by low-SES users.

Literacy/Culture/ Education H8: When technology uses complex language structures, advanced vocabulary, or cultural references, low-SES users will be more likely than higher-SES users to experience problems with the technology. H9: Technology that is culturally flexible (e.g., can be made to reflect the culture with which an individual most identifies) will be more inclusive across SES levels than technology that is not culturally flexible.

Control & Auhority H10: When low-SES users run into errors, dead-ends, or unfavorable outcomes, they will be more likely than higher-SES users to accept them as outcomes to which they have no recourse. H11: Low-SES users will be less likely than higher-SES users to engage in the kinds of planfulness or investing in the future needed to learn complex technological feature sets, or to plan a complex strategy for overcoming a problem with technology.

Risk, Privacy, and Security H12: Low-SES individuals will be less likely than high-SES individuals to successfully complete tasks with problem-solving software due to aversion to technological risks (e.g., aversion to potentially wasting time on new/unfamiliar features, aversion to potential loss of privacy, …). H13: Low-SES individuals will be more likely than high-SES individuals to perceive technology features as being risky, and therefore will be less likely to use those features.

leaves important phenomena undiscovered [143] . Intersectional HCI research is still in its infancy, so how to use the SES facets we have presented here in an intersectional approach remains open.

Utility also requires actionability, and we see actionability as a strength of the SES facet set, in combination with prior researchers' contributions relating to these facets. For example, the following actionable "next steps" are possible for HCI researchers and/or practitioners to add socioeconomic awareness to their own HCI con-

A socioeconomic-aware vocabulary: HCI practitioners can immediately use the SES facets as a vocabulary in design discussions. For example, instead of trying to ask and answer "will low-SES people use this feature?" they can use the vocabulary to focus on each facet one by one-e.g., "will people who are riskaverse use this feature?" A socioeconomic-aware design "starter set": Identifying SES-inclusivity problems using an SES-aware vocabulary leads to a need to fix those problems. Section 5's design recommendations provide ways to act upon this need. Because each of the design recommendations presented has at least some research or effectiveness data behind it, the recommendations provide a starting set of design remedies, organized by the SES facet/vocabulary that pinpointed the design aspects that need remedying. Socioeconomic-faceted personas: For researchers and practitioners who use personas, a possibility is to build socioeconomic-faceted personas, as in the InclusiveMag family [112] . To serve as an affordance for socioeconomic-inclusive design, at least two personas will be needed: one persona with facet values at the low end of each facet's value ranges, and a second persona at the high end. Additional personas with various mixes of the facet values may also be needed. Field insights from socioeconomic-aware HCI: We have pointed to the need for field research. Practitioners and researchers can gather field insights and data from teams working in any of the above ways to help socioeconomic awareness to grow and mature in HCI. For example, we have pointed out that the facets may need further modification, informed by evidence and insights from use in HCI practice. Hypothesis investigations: This paper has derived numerous hypotheses, and these offer opportunities for HCI researchers interested in doing empirical research to further socioeconomic awareness in HCI. Table  3 summarizes the open questions presented as hypotheses throughout Section 4.

Finally, we plan to follow the remainder of the InclusiveMag meta-method to build from these facets a systematic inclusive design process for socioeconomic awareness: SocioeconomicMag (Socioeconomic Inclusiveness Magnifier). We hope that, with this method, HCI practitioners will have the socioeconomic awareness they need to build user experiences that avoid SES-inclusivity problems like this one relating to the risk facet: "It's so scary to see that [enforcement] checks everything about you. I mean, you have no privacy... They could probably hack into my messages..." [68] ment; and all authors of the papers cited in the References section for their research and insights, which together will help make technology more inclusive. This work was supported in part by NSF grants 1901031 and 210045.

[ 

Overlap: This paper follows part of the InclusiveMag metamethod (our work, [112] ), and derives Figure 1 from that paper. Otherwise, this paper does not significantly overlap with the InclusiveMag paper or any other papers.

Prior submissions: A previous paper based on a subset of the references we use here, in the form of a systematic mapping paper, was submitted in the past but not accepted. We eventually decided that a systematic mapping was not the right approach, so we threw that paper away entirely, and started fresh to create the synthesis we are currently submitting. The current paper has never been submitted to any conference or journal. Table A enumerates the complete literature review we conducted. The number of papers covered for each facet does not reflect the number of papers that exist on that topic but rather the papers we covered via the process explained in Section 3, focusing on SES aspects with ties or potential impacts on HCI. In total, we covered over 200 papers, a subset of which are cited in this paper. Example: Low-SES Children -Most low-SES children wanted to play games, however, were limited because they either didn't own devices or had to share devices [9] . Example: Low-SES Teenagers -To encourage low-SES participation in classroom contexts, researchers developed an online tool to allow for reactions to others' assignments [116] Example: Low-SES Parents -One goal: to communicate with their family, however, shared devices pose a barrier to coordinating with others. Security issues also a concern with shared devices. [141, 188] Example: Low-SES Older Adults -Important to contact remote caregivers to stay connected, but many systems are unreliable [13] . [9, 11, 13, 17, 26, 40, 59, 85, 90, 103, 116, 127, 131, 132, 135, 141, 142, 144, 179, 182, 188] Culture Examples: -Virtual peer tutor helped low-to mid-SES African American students pick up "school English" while also allowing students to remain connected to own dialect and culture [131] -Latino parents have felt excluded from school communities (e.g., due to socioeconomic status, social history, English language mastery). Culturally responsive family-school partnerships may empower minority parents to advocate for their children's education and thereby to place hope in education as a means of social mobility [62] -African-American children of low-SES understand narrative information better than children of high SES [64] -Cultural knowledge stores (e.g., household practices, climate adaptations) cultural minority/low-SES groups already have potential resources from which teachers could draw to support students' literacy and content learning, schools with bilingual education programs tend to serve working-class students whose knowledge stores and potential/desire to succeed are often underestimated, further reducing educational quality [113, 173] 

Examples: -Asked about concerns and perceived risks of technology use, many undocumented immigrants mentioned security-related threats that are not unique to their immigration status-concerns such as identity theft, online financial fraud, unauthorized access to their Facebook accounts, and hackers who might steal their information or impersonate them online [68] -Overall, participants primarily managed privacy and security concerns through a limited set of practices: network regulation, varying degrees of self-censorship and deliberate choice of communication channels [68] . -Low-SES people concerned about their privacy, identity and even online harassment [68] -One privacy practice by low-SES individuals, beyond standard practices such as phone locks: avoiding many apps entirely, to maintain privacy while adhering to the cultural expectation that they should share their mobile phones with their social relations [145] -Another privacy practice by low-SES individuals: removing sensitive content through deleing entire threads or histories of content, or entity deletions, where participants deleted specific chats, media, or queries [145] [ 16, 68, 145, 167, 159] Race Examples: -On intersectionality: race is closely intertwined with SES trends, racial minorities disproportionately more likely to have low-SES characteristics (low income, low education, poor healthcare, etc.) [44] -Caucasians more likely to use computers and to have higher rates of internet access and broadband internet than African-Americans, who are more likely to use mobile phones and play/buy video games [188] [ 14, 31, 44, 188] Risk (Attitude toward)

Examples: -Low-SES people are exposed to greater technological risk than high-SES people. For example, undocumented Latinx immigrants are at risk of deportation/arrest if privacy is breached online [68] -Low-SES people (statistically) lower risk tolerance than high-SES people [67] -Low-SES people more likely to use ineffective emotion-focused coping strategies as opposed to problem-focused coping strategies. This greater difficulty in effectively solving problems when they arise further decreases risk tolerance. [ 

Gender and socioeconomic related differences in performance based ICT competences

Literature Review on the Feasibility of Mobile-Assisted Language Learning (MALL) in Developing Vocabulary Skills Among Non-English Speaking Migrant and Refugee Women

Social Distancing Challenges for Marginal Communities during COVID-19 Pandemic in Bangladesh

Socioeconomic Differences in Reading Trajectories: The Contribution of Family, Neighborhood, and School Contexts

Development of ICTL Module-Based on Suitable Teaching Methodology for Orang Asli (ICTL-TeMfOA): The Preliminary Study of Suitable Teaching Strategy

Vulnerability of Displacement: Challenges for Integrating Refugees and Asylum Seekers in Host Communities

How first-generation college and underrepresented students can overcome obstacles to attaining a college education: Handbook for a new family tradition

APA Dictionary of Psychology

Connected Learning" and the Equity Agenda: A Microsociology of Minecraft Play

Understanding Technology Choices and Values Through Social Class

Audience Involvement and Agency in Digital Games: Effects on Learning, Game Experience, and Social Presence

Breaking Barriers or Locked Out? Class-Based Perceptions and Experiences of Postsecondary Education

From Checking On To Checking In: Designing for Low Socio-Economic Status Older Adults

Family socioeconomic status at birth and youth impulsivity at age 15

Sessions with Grandma: Fostering Indigenous Knowledge through Video Mediated Communication

Americans and Privacy: Concerned, Confused and Feeling Lack of Control Over Their Personal Information

Like a Hustler: Aligning Intervention Design with Informal Labor Practices

Perceived Self-Efficacy in Cognitive Development and Functioning

Feminist HCI: Taking Stock and Outlining an Agenda for Design

Understanding the needs of low SES patients with type 2 diabetes

Reasoning about Many-to-Many Requirement Relationships in Spreadsheets

Experiences of students from diverse backgrounds: The role of academic support

Online Health Information and Low-Literacy African Americans

Divided We Call: Disparities in Access and Use of Mobile Phones in Rwanda. Information

Surveillance and system avoidance: Criminal justice contact and institutional attachment

Mobile Devices for Early Literacy Intervention and Research with Global Reach

An Intelligent Serious Game for Supporting African and African Caribbean Men during Pre-and Post-Diagnosis of Prostate Cancer

Finding gender-inclusiveness software issues with GenderMag: A field investigation

GenderMag: A Method for Evaluating Software's Gender Inclusiveness

The GenderMag Teach Project

Black Men in IT: Theorizing an Autoethnography of a Black Man's Journey into IT within the United States of America

Socioeconomic Status and Financial Coping Strategies: The Mediating Role of Perceived Control

I Need Help!" Social Class and Children's Help-Seeking in Elementary School

Racial and ethnic differences in advance care planning: Identifying subgroup patterns and obstacles

FLight: A Low-Cost Reading and Writing System for Economically Less-Privileged Visually-Impaired People Exploiting Ink-based Braille System

Evaluation of a Food Portion Size Estimation Interface for a Varying Literacy Population

Mobile interface design for low-literacy populations

Socioeconomic status and health: Mediating and moderating factors

Plain Language to Minimize Cognitive Load: A Social Justice Perspective

How to assign students into sections to raise learning

Accessing a New Land: Designing for a Social Conceptualisation of Access

Knowledge management in LLiSA ICT4D projects

Demographic predictors of senior secondary participation in biology, physics, chemistry and earth/space sciences: students' access to cultural, social and science capital

Demarginalizing the intersection of race and sex: A black feminist critique of antidiscrimination doctrine, feminist theory and antiracist politics

The Ins and Outs of HCI for Development

Social-ethnic school composition and disengagement: An inquiry into the perceived control explanation

Is literacy and the practice of literacy on the decline?

Effective teaching and support of students from low socioeconomic status backgrounds: Resources for Australian higher education

The Promise of the Sharing Economy among Disadvantaged Communities

Do Massive Open Online Course Platforms Support Employability?

Uncovering the Values and Constraints of Realtime Ridesharing for Low-resource Populations

Designing for Disadvantaged Job Seekers: Insights from Early Investigations

Achievement gap: Socioeconomic status affects reading development beyond language and cognition in children facing poverty

The digital divide at the margins: Co-designing information solutions to address the needs of indigenous populations of rural India

Development in Early Childhood

Engaging Around Neighborhood Issues: How Online Communication Affects Offline Behavior

Differences in technology use to support community crime prevention

Computer and Internet Use in the United States

Action!: codesigning interactive technology with immigrant teens

Seeking Structure in Social Organization: Compensatory Control and the Psychological Advantages of Hierarchy

Income and Poverty in The United States

Education inequalities at the school starting gate: Gaps, trends, and strategies to address them

Financial Risk Tolerance and Additional Factors That Affect Risk Taking in Everyday Money Matters

Keeping a Low Profile?: Technology, Risk and Privacy Among Undocumented Immigrants

How To Debug Inclusivity Bugs? An Empirical Investigation of Finding-to-Fixing with Information Architecture

Non-Native English Speakers Learning Computer Programming: Barriers, Desires, and Design Opportunities

Placating the Powerless: Effects of Legitimate and Illegitimate Explanation on Affect, Memory, and Stereotyping

Moving beyond the study of gender differences: An analysis of measurement invariance and differential item functioning of an ICT literacy scale

2016. mBody Health: Digitizing Disabilities in Sierra Leone

Engineering Gender-Inclusivity into Software: Tales from the Trenches

General Self-Efficacy Mediates the Effect of Family Socioeconomic Status on Critical Thinking in Chinese Medical Students

Status, Religion, and the

Harm in Asking: Class, Acquired Cultural Capital, and Academic Engagement at an Elite University

Don't blame me, it's just the computer telling me to do this": Computer Attribution and theDdiscretionary Authority of Canada Border Services Agency Officers

Transforming Traditional Pervasive Computing for Emerging Markets -Underlying Challenges and Opportunities

God and the government: Testing a compensatory control mechanism for the support of external systems

Socioeconomic Status and Belief in a Just World

Designing Mobile Snack Application for Low Socioeconomic Status Families

Exploring Everyday Health Routines of a Low Socioeconomic Population Through Multimedia Elicitations

Towards Equal Opportunities in MOOCs: Affirmation Reduces Gender & Social-Class Achievement Gaps in China

Computing Mentorship in a Software Boomtown: Relationships to Adolescent Interest and Beliefs

Social Class and the Exercise of Parental Authority

Mobile Phones for Maternal Health in Rural India

Literacy in Everyday Life: Results from the 2003 National Assessment of Adult Literacy. NCES 2007-490

Ceci N'est Pas Une Pipe Bombe: Authoring Urban Landscapes With Air Quality Sensors

Understanding Intersecting Social Identities in Engineering Education and Practice

Restricted Emigration, System Inescapability, and Defense of the Status Quo: System-Justifying Consequences of Restricted Exit Opportunities

Social Disadvantage and the Self-Regulatory Function of Justice Beliefs

Designs on Dignity: Perceptions of Technology Among the Homeless

Personifying Programming Tool Feedback Improve Novice Programmers' Learning

Technologies in Use for Second Language Learning

Investigating Country Differences in Mobile App User Behavior and Challenges for Software Engineering

How socio-economic structure influences rural users' acceptance of mobile entertainment

Improving Communication and Social Support for Caregivers of High-Risk Infants Through Mobile Technologies

Large Dataset Offers View of Math and Computer Self-Efficacy Among Computer Science Undergraduates

Rural Teachers' Literacy Practices In and Out of the Classroom: Exploring Teacher Characteristics and Literacy Tools

Print Literacy Engagement of Parents From Low-Income Backgrounds: Implications for Adult and Family Literacy Programs

Persuasion Not Required: Improving Our Understanding of the Sociotechnical Context of Dietary Behavioural Change

The Importance of Cooperation Centres for HE Students' Dropout Prevention

Stuck in the Shallow End

Mobile Phones and Inequality: Findings, Trends, and Future Directions

Path Not Found: Disparities in Access to Computer Science Courses in California High Schools. ERIC No. ED561181

Nobody Sees It, Nobody Gets Mad: Social Media, Privacy, and Personal Responsibility Among Low-SES Youth

Jury: An Automation Framework for Protocolised Primary Healthcare Delivery

Is the Belief in Meritocracy Palliative for Members of Low Status Groups? Evidence for a Benefit for Self-Esteem and Physical Health via Perceived Control

Coping with Occupational Problems: The Limits of Individual Efforts

The InclusiveMag Method: A Start Towards More Inclusive Software for Diverse Populations

From GenderMag to InclusiveMag: An Inclusive Design Meta-Method

Bilingual Classroom Studies and Community Analysis: Some Recent Trends

Emergent Practices Around CGNet Swara, Voice Forum for Citizen Journalism in Rural India

Low-Income Parents' Perceptions of Technology: Value-Based Design Insights

Encouraging Online Contributions in Underrepresented Populations

5G Network Slicing for Digital Inclusion

Multimodal Approaches to Raising Bilingual

Kenya Digital and Cultural Divide

How Technology Supports Family Communication in Rural, Suburban, and Urban Kenya

Identifying Pathways Between Socioeconomic Status and Language Development

How Gender-Biased Tools Shape Newcomer Experiences in OSS Projects

Perceived Control and Mindfulness: Implications for Clinical Practice

Marginality, Aspiration and Accessibility in ICTD

Design Principles for an Adult Literacy Mobile Learning Solution

World Conference on Mobile and Contextual Learning

The Multi-layers of Digital Exclusion in Rural Australia

A Critical Research Synthesis of Privilege in Computing Education

2020. Poverty, Inequality and COVID-19: The Forgotten Vulnerable

Avaaj Otalo: A Field Study of an Interactive Voice Forum for Small Farmers in Rural India

Out of Luck": Socio-Economic Differences in Student Coping Responses to Technology Problems

Brick By Brick: Iterating Interventions to Bridge the Achievement Gap with Virtual Peers

Exploring iPhone Usage: The Influence of Socioeconomic Differences on Smartphone Adoption, Usage and Usability

From Infomediaries to Infomediation at Public Access Venues: Lessons from a 3-Country Study

African American Single Mothers: Understanding Their Lives and Families

Anthropology, Development and ICTs: Slums, Youth And The Mobile Internet in Urban India

Five Evils: Multidimensional Poverty and Race in America. Economic Studies at Brookings Report

Development and Validation of the Parental Authority Questionnaire -Revised

Blind Spots in Youth DIY Programming: Examining Diversity in Creators, Content, and Comments within the Scratch Online Community

Opportunity For All? Technology and Learning in Lower-Income Families. A Report of the Families and Media Project

At Home with Agents: Exploring Attitudes Towards Future Smart Energy Infrastructures

Exploring How Parents in Economically Depressed Communities Access Learning Resources

Designing For Repair?: Infrastructures and Materialities of Breakdown

The Intersection of Being Black and Being a Woman: Examining the Effect of Social Computing Relationships on Computer Science Career Choice

Student and Faculty Inter-Generational Digital Divide: Fact or Fiction?

Privacy is not for me, it's for those rich women": Performative Privacy Practices on Mobile Phones by Women in South Asia

Designing a Phone Broadcasting System for Urban Sex Workers in India

Investigating Academic Self-Efficacy of University Students in Terms of Socio-Demographic Variables

Intersectional HCI: Engaging Identity through Gender, Race, and Class

Empowerment in HCI -A Survey and Framework

Sarah Newcomer, and Don Mcmahon. 2020. Second Language Socialization in Stronghold Kingdoms: A Narrative Inquiry

Towards a Design Model for Women's Empowerment in the Developing World

Children of Color's Perceptions of Fairness in AI: An Exploration of Equitable and Inclusive Co-Design

Digital Divide: Students' Use of the Internet and Emerging Forms of Social Inequalities

Moving Beyond Digital Literacy and Towards e-Governance in Rural India

Supporting Community Needs for Rural Water Management Through Community-Based Co-Design

Designing and Evaluating mHealth Interventions for Vulnerable Populations: A Systematic Review

Exploring Learning Ecologies Among People Experiencing Homelessness

Gender-Inclusive HCI Research and Design: A Conceptual Review. Foundations and Trends in Human-Computer Interaction

Online Privacy Fears are Real, NBC News

Informing EMR System Design through Investigation of Paper-Based Work Practices in a Non-profit Clinic Serving a Vulnerable Population

Diversity and Coherence in a Hackerspace for People from a Low Socioeconomic Community

Avoiding the South Side and the Suburbs: The Geography of Mobile Crowdsourcing Markets

Resource or Obstacle?: Classed Reports of Student-Faculty Relations

Family Socioeconomic Status and Adolescent Attitudes to Authority

Domains and Determinants of University Students' Self-Perceived Computer Competence

A Sense of Powerlessness Fosters System Justification: Implications for the Legitimation of Authority, Hierarchy, and Government

Mind the five": Guidelines for Data Privacy and Security in Humanitarian Work with Undocumented Migrants and Other Vulnerable Populations

Educational Content Creation and Sharing by Low-Income Visually Impaired People in India

Social Media Platforms for Low-Income Blind People in India

Examining Localization Approaches for Community Health

Socioeconomic Differences in Elementary Students' ICT Beliefs and Out-of-School Experiences

Gender issues in technology use: Perceived social support, computer self-efficacy and value beliefs, and computer use beyond school

Formation and Transformation of Funds of Knowledge Among U.S.-Mexican Households

Pay or Delay: The Role of Technology When Managing a Low Income

Gender Biases to Gender-Inclusive Design: An Empirical Investigation

More Than Just Food: Field Visits to an Emergency Relief Centre

Citizen Involvement in the Design of Technology for Climate Change Adaptation Projects in The Pacific

Landscape of K-12 Computer Science Education in the U.S.: Perceptions, Access, and Barriers

Digital Divides? Student and Staff Perceptions of Information and Communication Technologies

The Cognitive Walkthrough Method: A Practitioner's Guide

Belief in School Meritocracy as a System-Justifying Tool for Low Status Students

Exploring Student Motivation Towards Diversity Education in Engineering

Improving the Safety of Homeless Young People with Mobile Phones: Values, Form and Function

Exploring mobile phone and social media use in a Nairobi slum: a case for alternative approaches to design in ICTD

Dead China-make" Phones off the Grid: Investigating and Designing for Mobile Phone Use in Rural Africa

Revisiting the status-legitimacy hypothesis: Concepts, boundary conditions, and psychological mechanisms

Explaining the Global Digital Divide: The Impact of Public Policy Initiatives on Digital Opportunity and ICT Development

Income, Race, and Class: Exploring Socioeconomic Differences in Family Technology Use

Parenting in Families of Low Socioeconomic Status: A Review with Implications for Child Welfare Practice

Low-SES families spend more time and energy than higher-SES families feeding, clothing, and caring for their children. Their time also goes to complicated logistics due to use of public transportation and reliance on bureaucracies for income

We thank those who have provided expertise on socioeconomic phenomena through their own life experiences and through their research expertise; Heather DiRuscio and Caleb Matthews for help in preparing this docu-