<scp>DARPA</scp>'s explainable<scp>AI</scp>(<scp>XAI</scp>) program: A retrospective

Gunning, Dave; Vorm, Eric S.; Wang, Yunyan; Turek, Matt

doi:10.1002/ail2.61

Cited by 102 publications

(35 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Incorporating these factors will help us better understand the impact of consensus-based explanations on trust and reliance. Some subjects were unable to use some data source tweet (5,8) type of consensus the agent provided. the same. "…”

Section: Resultsmentioning

confidence: 99%

See 1 more Smart Citation

Trust and Reliance in Consensus-Based Explanations from an Anti-Misinformation Agent

Ueno

Kim

Oura

et al. 2023

Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

The illusion of consensus occurs when people believe there is consensus across multiple sources, but the sources are the same and thus there is no "true" consensus. We explore this phenomenon in the context of an AI-based intelligent agent designed to augment metacognition on social media. Misinformation, especially on platforms like Twitter, is a global problem for which there is currently no good solution. As an explainable AI (XAI) system, the agent provides explanations for its decisions on the misinformed nature of social media content. In this late-breaking study, we explored the roles of trust (attitude) and reliance (behaviour) as key elements of XAI user experience (UX) and whether these influenced the illusion of consensus. Findings show no effect of trust, but an effect of reliance on consensus-based explanations. This work may guide the design of anti-misinformation systems that use XAI, especially the user-centred design of explanations.

show abstract

Section: Resultsmentioning

confidence: 99%

“…To address these issues, the notion of explainable AI (XAI) has gained momentum in recent years. The purpose of XAI, a term coined by DARPA researchers in 2017, is to make AI decision-making and other fundamental behavior more understandable to people by providing a human-parsable explanations [8].…”

Section: Introductionmentioning

confidence: 99%

Trust and Reliance in Consensus-Based Explanations from an Anti-Misinformation Agent

Ueno

Kim

Oura

et al. 2023

Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems

View full text Add to dashboard Cite

show abstract

“…Проблема черного ящика Основная трудность с применением требования разъяснения состоит в том, что многие системы ИИ работают как "черные ящики": они могут быть искусными в выполнении задач, но даже их собственные конструкторы могут быть не в состоянии объяснить, каким образом внутреннии� процесс привел к определенному результату [15].…”

Section: требование четкого разъясненияunclassified

Artificial intelligence – WHEther responsibility of robots possible?

Bozarov¹

2021

jurisprudence

View full text Add to dashboard Cite

In this scientific article, the author examines the existing scientific and theoretical approaches to determine the legal responsibility of artificial intelligence, in particular, robots, which are currently the most prominent representatives of the field of artificial intelligence. As the author notes, the rules and mechanisms for establishing the legal responsibility of robots will change and improve over time, however, a clear regulation of the set of social principles - the rules of the “life” of robots, is required. In the article, the author provides a list of similar principles such as the requirement for identification of robots, the requirement for a warning, the “black box problem”, the requirement for the independence of artificial intelligence from the engineering infrastructure and human factor, the requirement for the usefulness of robotization, the requirement for punishment (removal), the requirement of correct ability and the problem of “turning off” robots. As the main problems of the legal regulation of artificial intelligence, the author points out the autonomy and independence of the creation of artificial intelligence objects, which are now being produced on an industrial scale in many countries. At the same time, the lack of single principles for their creation and operation seriously complicates both the unification of requirements for them and the determination of measures of legal responsibility. In conclusion, it is noted that with a further increase in the use of robots, a balance should be found between the usefulness of robotization and the safety of society.

show abstract

“…In general, there may be several minimal explanations for a particular prediction. Depending on the application domain and target audience, different explanations may be preferable (Gunning et al, 2021). In the medical domain for example, some explanations may require that the target audience has medical training in order to comprehend concepts captured by variables associated with an explanation, while other explanations may be more suitable for patients, albeit more verbose.…”

Section: Definition 2 (Minimal Explanationmentioning

confidence: 99%