When less is more: single selfhood-related cues elicit higher selfhood ratings than multiple cues

In everyday social cognition, people reliably tend to anthropomorphize non-human agents or ascribe to them features associated with selfhood. In prior research, we showed that individuals often (over-)generalize from a single behavioral indicator of selfhood, such as learning ability or attentional sharing, to the assumption that other, unobserved indicators of selfhood are also present. This suggests that even minimal cues may be sufficient to activate a broader selfhood concept and its associated implications.

Here, we tested the prediction that single selfhood cues are as effective as multiple selfhood cues in eliciting selfhood- attributions to an artificial agent. In three experiments, we compared selfhood ratings for a robot exhibiting behavioral cues of efficiency, learning sensitivity, and equifinality with ratings for a robot exhibiting only one of these cues.

Contrary to our expectations, participants did not attribute the same degree of selfhood to both robots. They attributed more selfhood to the single-cue robot in Experiments 1 (single efficiency cue) and 3 (single equifinality cue), whereas in Experiment 2 (single learning cue), the multiple-cue robot received higher ratings on the manipulated characteristics as well as on context sensitivity (results for selfhood-attribution were negligible).

In sum, we conclude that a single selfhood cue tends to elicit stronger selfhood-attributions overall than multiple cues. We discuss this finding in relation to the literature on cognitive bootstrapping (i.e., developing new and complex knowledge from a limited starting set), concluding that consistently applying a single skill to an apparent problem may reflect more human-like behavior.