Explainable Artificial Intelligence

Algorithmic Literacy and the Role for Libraries

“…Perhaps a less obvious way for libraries to contribute to algorithmic literacy is through explainable AI (XAI). 61 Difficulties in interrogating algorithms to assess bias, discrimination, and unfairness (as well as other deficiencies such as veracity and generalizability) have led to widespread interest in XAI. The purpose of XAI is to "enable human users to understand, appropriately trust, and effectively manage the emerging generation of artificially intelligent partners" and to deploy AI systems that have "the ability to explain their rationale, characterize their strengths and weaknesses, and convey an understanding of how they will behave in the future."…”

Artificial Intelligence in Legal Practice: Exploring Theoretical Frameworks for Algorithmic Literacy in the Legal Information Profession

“…The responsible incorporation and adoption of artificially intelligent systems may be assisted by meeting the opacity of such systems with initiatives aimed at explainability and user inclusivity. Ridley (2019) 69 , citing de Mul and van den Berg (2011) 70 , acknowledges that the danger of reliance on artificially intelligent systems is not so much in the increased delegation of cognitive tasks to these systems, but in information professionals and information users distancing themselves from, and not knowing about, the nature, precise mechanisms, and repercussions of that delegation. Adding to this, Henry (2019) 71 states that instilling policies that require accountability, mandating not just access to the algorithms themselves and the processes followed when using the data but an accessible explanation of the extent to which the data used, is a key aspect of future governance and regulatory frameworks that foster ethically responsible behaviors in the use of intelligent systems.…”

“…Adding to this, Henry (2019) 71 states that instilling policies that require accountability, mandating not just access to the algorithms themselves and the processes followed when using the data but an accessible explanation of the extent to which the data used, is a key aspect of future governance and regulatory frameworks that foster ethically responsible behaviors in the use of intelligent systems. Ridley (2019) 72 explores the concept of explainable artificial intelligence, which is broadly defined as a diverse set of strategies, techniques, and processes that render artificially intelligent systems interpretable and accountable. Trust and accountability are deemed the two pillars of explainable artificial intelligence, and Ridley emphasizes user-centred explainability as an essential requirement for a technology defined by its opacity.…”

“…Teaching and learning initiatives grounded on approaches such as reviewing outcomes of reverse engineering tests, interviewing designers or conducting ethnography studies of coding teams, unpacking the sociotechnical assemblages of algorithms, and examining how algorithms work in the world, along with elements such as human computer interaction and information seeking behaviour, all facilitate increasing computer literacy and algorithmic awareness in the use of artificially intelligent systems in legal information and research contexts. These 83 , who states that some approaches of explainable artificial intelligence may involve studies on broad social and political policies surrounding regulation, legislation, or practice, without necessarily requiring highly technical expertise.…”

“…Swansburg (2021) 13 notes how advanced legal technologies are an enigma for most practitioners, as it is challenging to possess a clear sense of how the information, answers, and work product of artificially intelligent tools are created. Ethical issues emerging from decreased human oversight and increased reliance on these 'black box' technologies have been discussed by other authors such as Callister (2020) 14 , Nayyer et al (2020) 15 , Wojcik (2020) 16 , Yu and Ali (2019) 17 , Bhattacharya et al (2019) 18 , Gardner (2019) 19 , Henry (2019) 20 , Ridley (2019) 21 , Turner (2018) 22 , Baker (2018a) 23 , and Kennedy (2017) 24 .…”