explainable artificial intelligence ( xai ) adoption and advocacy michael ridle
explainable artificial intelligence ( xai ) advances techniques , processes , a
ing . academic libraries should adopt xai as a tool set to verify and validate 
 advocate for public policy regarding xai that serves libraries , the academy ,
explainable artificial intelligence ( xai ) is a subfield of artificial intelli
integral part of academic libraries . xai is a set of techniques , processes , 
only recently turned its attention to xai and how it affects the field and how 
nd how the field might influence it.5 xai is a critical lens through which to v
mportant.6 dismissing engagement with xai because it is “ highly technical and 
 contemporary information systems and xai is a tool set for critical assessment
s are to have a place at the table as xai , and machine learning , evolves and 
explainable artificial intelligence ( xai ) | ridley 2 this paper provides an o
 2 this paper provides an overview of xai with key definitions , a historical c
 historical context , and examples of xai techniques , strategies , and process
 the field . it considers areas where xai and academic libraries intersect . th
s intersect . the dual emphasis is on xai as a toolset for libraries to adopt a
 a toolset for libraries to adopt and xai as an area for public policy advocacy
 for public policy advocacy . what is xai ? xai is plagued by definitional prob
ublic policy advocacy . what is xai ? xai is plagued by definitional problems.8
 systems.11 as such , a definition of xai must encompass not just the technique
e , but also the context within which xai operates . the us defense advanced re
jects agency ( darpa ) description of xai captures the breadth and scope of the
d scope of the field . the purpose of xai is for ai systems to have “ the abili
tificially intelligent partners. ” 13 xai is needed to : 1. generate trust , tr
date explanations generated by xai.14 xai consists of testable and unambiguous 
ng within a public policy framework . xai is not a new consideration . explaina
 evitability spurring an explosion in xai research and development.18 types of 
 research and development.18 types of xai taxonomies of xai types are classifie
lopment.18 types of xai taxonomies of xai types are classified based on their s
explainable artificial intelligence ( xai ) | ridley 3 white-box or model-speci
the model . another way to categorize xai is as proofs , validations , and auth
ing . libraries must also engage with xai as authorizations to assess the publi
l of people. ” 23 prerequisites to an xai strategy three questions are importan
three questions are important for any xai strategy : • what constitutes a good 
nee ( a user ) and the explainer ( an xai ) . following research from the field
explainable artificial intelligence ( xai ) | ridley 4 a good explanation is al
teracy . second is a requirement that xai must be sensitive to the abilities an
 challenges and research direction of xai identified 39 issues , including the 
d enhance the user experience , match xai to user expertise , and explain the c
 approaches.33 proofs and validations xai that provide proofs or validations ca
d abstraction , reproducibility , and xai by ai . these techniques may require 
explainable artificial intelligence ( xai ) | ridley 5 available , designers of
explainable artificial intelligence ( xai ) | ridley 6 generalizability of thes
rtant to this type of verification.48 xai by ai the inherent complexity and opa
 reinforcement learning suggests , as xai researcher trevor darrell puts it , “
aries and the academy . authorization xai that results from authorizations is a
policy engagement is needed to ensure xai , and machine learning , are appropri
explainable artificial intelligence ( xai ) | ridley 7 unfortunately , the opti
explainable artificial intelligence ( xai ) | ridley 8 systems will remain blac
 any such auditing responsibility for xai would require the trust of stakeholde
sly assess an algorithmic system ; an xai form of the “ secret shopper ” ) .74 
the rise of public activists into the xai arena . the complexity of algorithms 
opriate for scholarly communication . xai as discovery while xai is primarily a
ommunication . xai as discovery while xai is primarily a means to validate and 
ine learning systems , another use of xai is gaining attention . since xai can 
e of xai is gaining attention . since xai can find new information latent in la
merging reason for libraries to adopt xai may be as a powerful discovery tool .
explainable artificial intelligence ( xai ) | ridley 9 conclusion our lives hav
acterises the presence of trust. ” 85 xai is an essential tool to build that tr
norms of algorithmic accountability . xai is a dual opportunity for libraries .
 . many disciplines have engaged with xai as machine learning has impacted thei
learning has impacted their fields.89 xai has been called a “ disruptive force 
arranting the growing interest in how xai affects the field and how the field m
explainable artificial intelligence ( xai ) | ridley 10 endnotes 1 vijay arya e
esina schwalbe and bettina finzel , “ xai method properties : a ( meta- ) study
explainable artificial intelligence ( xai ) , ” records management journal ( lo
explainable artificial intelligence ( xai ) | ridley 11 landscape-summary.pdf ;
 151 . 8 sebastian palacio et al. , “ xai handbook : towards a unified framewor
explainable artificial intelligence ( xai ) ( arlington , va : darpa , 2016 ) ,
explainable artificial intelligence ( xai ) , ” darpa , https : //www.darpa.mil
explainable artificial intelligence ( xai ) , ” in proceedings of the hawaii in
explainable artificial intelligence ( xai ) | ridley 12 17 lilian edwards and m
explainable artificial intelligence ( xai ) , ” ieee access 6 ( 2018 ) : 52138–
ligence. ” 19 schwalbe and finzel , “ xai method properties. ” 20 or biran and 
explainable artificial intelligence ( xai ) , melbourne , 2017 ) , http : //www
explainable artificial intelligence ( xai ) | ridley 13 28 tim miller , “ expla
 christian omlin , “ explainable ai ( xai ) : a systematic meta-survey of curre
ijcai-17 workshop on explainable ai ( xai ) , melbourne : ijcai , 2017 ) , http
explainable artificial intelligence ( xai ) | ridley 14 40 tania lombrozo , “ e
explainable artificial intelligence ( xai ) | ridley 15 york : ai now institute
explainable artificial intelligence ( xai ) | ridley 16 66 danielle keats citro
explainable artificial intelligence ( xai ) | ridley 17 81 ahmed alkhateeb , “ 
nce.htm abstract introduction what is xai ? types of xai prerequisites to an xa
t introduction what is xai ? types of xai prerequisites to an xai strategy proo
ai ? types of xai prerequisites to an xai strategy proofs and validations featu
ation and abstraction reproducibility xai by ai authorization codes and standar
 codes and standards regulation audit xai as discovery conclusion endnotes lett
explainable artificial intelligence ( xai ) : adoption and advocacy / michael r