TY - JOUR
T1 - Moral empiricism and the bias for act-based rules
AU - Ayars, Alisabeth
AU - Nichols, Shaun
N1 - Funding Information:
We would like to thank Victor Kumar, Fiery Cushman, and three anonymous referees for extensive comments on an earlier version of this paper. Research for this paper was supported by Office of Naval Research grant #11492159 to SN.
Publisher Copyright:
© 2017 Elsevier B.V.
PY - 2017/10
Y1 - 2017/10
N2 - Previous studies on rule learning show a bias in favor of act-based rules, which prohibit intentionally producing an outcome but not merely allowing the outcome. Nichols, Kumar, Lopez, Ayars, and Chan (2016) found that exposure to a single sample violation in which an agent intentionally causes the outcome was sufficient for participants to infer that the rule was act-based. One explanation is that people have an innate bias to think rules are act-based. We suggest an alternative empiricist account: since most rules that people learn are act-based, people form an overhypothesis (Goodman, 1955) that rules are typically act-based. We report three studies that indicate that people can use information about violations to form overhypotheses about rules. In study 1, participants learned either three “consequence-based” rules that prohibited allowing an outcome or three “act-based” rules that prohibiting producing the outcome; in a subsequent learning task, we found that participants who had learned three consequence-based rules were more likely to think that the new rule prohibited allowing an outcome. In study 2, we presented participants with either 1 consequence-based rule or 3 consequence-based rules, and we found that those exposed to 3 such rules were more likely to think that a new rule was also consequence based. Thus, in both studies, it seems that learning 3 consequence-based rules generates an overhypothesis to expect new rules to be consequence-based. In a final study, we used a more subtle manipulation. We exposed participants to examples act-based or accident-based (strict liability) laws and then had them learn a novel rule. We found that participants who were exposed to the accident-based laws were more likely to think a new rule was accident-based. The fact that participants’ bias for act-based rules can be shaped by evidence from other rules supports the idea that the bias for act-based rules might be acquired as an overhypothesis from the preponderance of act-based rules.
AB - Previous studies on rule learning show a bias in favor of act-based rules, which prohibit intentionally producing an outcome but not merely allowing the outcome. Nichols, Kumar, Lopez, Ayars, and Chan (2016) found that exposure to a single sample violation in which an agent intentionally causes the outcome was sufficient for participants to infer that the rule was act-based. One explanation is that people have an innate bias to think rules are act-based. We suggest an alternative empiricist account: since most rules that people learn are act-based, people form an overhypothesis (Goodman, 1955) that rules are typically act-based. We report three studies that indicate that people can use information about violations to form overhypotheses about rules. In study 1, participants learned either three “consequence-based” rules that prohibited allowing an outcome or three “act-based” rules that prohibiting producing the outcome; in a subsequent learning task, we found that participants who had learned three consequence-based rules were more likely to think that the new rule prohibited allowing an outcome. In study 2, we presented participants with either 1 consequence-based rule or 3 consequence-based rules, and we found that those exposed to 3 such rules were more likely to think that a new rule was also consequence based. Thus, in both studies, it seems that learning 3 consequence-based rules generates an overhypothesis to expect new rules to be consequence-based. In a final study, we used a more subtle manipulation. We exposed participants to examples act-based or accident-based (strict liability) laws and then had them learn a novel rule. We found that participants who were exposed to the accident-based laws were more likely to think a new rule was accident-based. The fact that participants’ bias for act-based rules can be shaped by evidence from other rules supports the idea that the bias for act-based rules might be acquired as an overhypothesis from the preponderance of act-based rules.
KW - Act/allow distinction
KW - Bayesian learning
KW - Moral rules
KW - Overhypotheses
UR - http://www.scopus.com/inward/record.url?scp=85011019229&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85011019229&partnerID=8YFLogxK
U2 - 10.1016/j.cognition.2017.01.007
DO - 10.1016/j.cognition.2017.01.007
M3 - Article
C2 - 28159315
AN - SCOPUS:85011019229
SN - 0010-0277
VL - 167
SP - 11
EP - 24
JO - Cognition
JF - Cognition
ER -