The importance of epistemic intentions in ascription of responsibility
Abstract
We investigate how people ascribe responsibility to an agent who caused a bad outcome but did not know he would. The psychological processes for making such judgments, we argue, involve finding a counterfactual in which some minimally benevolent intention initiates a course of events that leads to a better outcome than the actual one. We hypothesize that such counterfactuals can include, when relevant, epistemic intentions. With four vignette studies, we show that people consider epistemic intentions when ascribing responsibility for a bad outcome. We further investigate which epistemic intentions people are likely to consider when building counterfactuals for responsibility ascription. We find that, when an agent did not predict a bad outcome, people ascribe responsibility depending on the reasons behind the agents’ lack of knowledge. People judge agents responsible for the bad outcome they caused when they could have easily predicted the consequences of their actions but did not care to acquire the relevant information. However, when this information was hard to acquire, people are less likely to judge them responsible.
Study specs
- Discipline
- Psychology,Behavioral Science
- Year
- 2024
- Human Data Platform
- Prolific
- Source
- View Source Google Scholar
Peer Review & Critical Discussion
Potential Selection Bias in 2023 Cohort
The participant pool shows a concerning overrepresentation of users from high-income demographics. Looking at Table 3, we can see that 78% of respondents had annual incomes above $75k, which significantly limits the generalizability of these findings to broader populations.
Non-naive Participants Issue
I've noticed a methodological concern regarding participant naivety. Given that Prolific users often complete multiple studies, there's a real risk that participants had prior exposure to similar experimental paradigms, which could confound the results.
RLHF Applicability to This Study Design
The implications for RLHF training pipelines are understated. If we accept the authors' conclusions about preference stability, this has direct consequences for how we should structure reward model training. The temporal decay effect described in Section 4.2 is particularly relevant.
Verify your expertise to join discussion
Create an account and verify your credentials to participate in peer discussions.