Algorithm aversion: people erroneously avoid algorithms after seeing them err.

By Berkeley J. Dietvorst, Joseph P. Simmons, Cade Massey. 2015

2,547 citations recorded by the source corpus.

Subfields: General Decision Sciences, Management Science and Operations Research, Sociology and Political Science, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In General Decision Sciences

In Management Science and Operations Research

In Sociology and Political Science

In Evaluation & Benchmarks

Citations in the corpus

Anticipating Safety Issues in E2E Conversational AI: Framework and Tooling (2021, background): “This effect is especially pronounced in algorithmic systems, where people apply higher standards than in their interaction with other humans (Dietvorst et al., 2015).”
Guiding the Release of Safer E2E Conversational AI through Value Sensitive Design (2022, background): “This effect is especially pronounced in algorithmic systems, where people apply higher standards than in their interaction with other humans (Dietvorst et al., 2015).”
Explainable Recommendation: A Survey and New Perspectives (2018, background): “The system also needs to be especially careful not to provide obviously misleading or even wrong explanations, because people may lose confidence in algorithms after seeing them trying to mistakenly convince others (Dietvorst et al., 2015).”
Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims (2020, background): “, studying the efficacy of interpretability tools or exploring possible interfaces–would be welcome, as would further exploration of how practitioners currently use such tools [81] [82] [83] [78] [84].”
Manipulating and Measuring Model Interpretability (2018, methodology): “racy comparisons since there was no ground truth. These apartments were always shown last to avoid the previously studied phenomenon in which people trust a model less after seeing it make a mistake (Dietvorst et al., 2015). 2.2 Results Having run our experiment, we compared participants’ behavior a”
Interpreting Blackbox Models via Model Extraction (2017, background): “…unexpected defects when deployed in the real world; examples include causality (i.e., inability to distinguish causal effects from correlations) [34, 16], fairness (i.e., internalizing prejudices present in training data) [22, 26], and algorithm aversion (i.e., lack of trust by end users) [19].”
Artificial intelligence versus Maya Angelou: Experimental evidence that people cannot differentiate AI-generated from human-written poetry (2020, background): “This reluctance of “ human decision makers to use superior but imperfect algorithms ” (Burton, Stein, & Jensen, 2019, p.1) has been referred to as algorithm aversion (Dietvorst, Simmons, & Massey, 2015).”
The Impact of Algorithms on Judicial Discretion : Evidence from Regression Discontinuities (2018, background): “Human experts are unwilling to defer to algorithms in many settings (Dietvorst et al., 2015b,a; Hoffman et al., 2015; Yeomans et al., 2017; Logg, 2017; Stevenson, 2017a); this fact is often problematic for human-computer interaction designers.”
Understanding perception of algorithmic decisions: Fairness, trust, and emotion in response to algorithmic management (2018, background): “Other research suggests that people trust their own judgment more, especially after the algorithm has erred, probably because they believe algorithms cannot learn from their mistakes (Dietvorst et al., 2015).”
Explaining Decision-Making Algorithms through UI: Strategies to Help Non-Expert Stakeholders (2019, background): “For instance, studies have shown that even when algorithmic predictions are proved to be more accurate than human predictions, both domain experts and laypeople remain resistant to using the algorithms (e.g. [17]).”
Explanations Can Reduce Overreliance on AI Systems During Decision-Making (2022, background): “As participants complete a series of maze solving tasks, we do not provide immediate feedback since humans are particularly sensitive to seeing algorithms err [19].”