Algorithm Aversion is Motivated: Symmetrical Partisan Bias in the Acceptance of Algorithmic Policy Advisors

We examine the influence of political orientation on attitudes toward Artificial Intelligence in policy-making. Liberals (vs. conservatives) are not necessarily more or less favourable to AI: They assess the desirability of AI based on whether it could disrupt a political status quo that aligns—or conflicts—with their beliefs. Specifically, liberals show higher acceptance of AI in policy-making, and also find AI more impartial, when the policy in question is conservative- (vs liberal-) leaning. Symmetrically, conservatives show higher acceptance of AI, and find it more impartial, when the policy is liberal- (vs. conservative-) leaning. These findings seem specific to AI, as they do not generalize to other policy advisors such as academics or consultants. This research provides an explanation (based on motivated reasoning) for past inconsistent findings on the relationship between political orientation and algorithm aversion, and also provides further evidence supporting the “symmetrical partisan bias” hypothesis in political psychology.