Preference for human, not algorithm aversion.

Area of Science:

• Cognitive Psychology
• Behavioral Economics
• Human-Computer Interaction

Background:

• A costly preference for human decision-makers over algorithms is often termed algorithm aversion.
• This aversion suggests a domain-general distrust of algorithms.
• However, the underlying psychological mechanisms remain incompletely understood.

Purpose of the Study:

• To challenge the notion of domain-general algorithm aversion.
• To propose an alternative explanation rooted in biased social cognition.
• To identify specific conditions under which human preference over algorithms is amplified.

Main Methods:

• The study likely involved experimental designs presenting participants with choices involving human versus algorithmic decision-makers.
• Evaluations of self and others were assessed, alongside domain-specific factors like identity threat and ambiguity.
• Statistical analyses were used to correlate these factors with preferences.

Main Results:

• Findings indicate that human preference is not domain-general but context-specific.
• Biased self- and other-evaluations significantly predict the preference for humans.
• This effect is pronounced in domains where personal identity is salient or evaluative standards are unclear.

Conclusions:

• Algorithm aversion is better explained by domain-specific biases in social evaluation rather than a general distrust of technology.
• Identity threat and ambiguous criteria are key moderators of the human-preference effect.
• Understanding these biases is crucial for designing human-AI interactions that mitigate unfair discrimination.