How peer influence shapes value computation in moral decision-making

Abstract

Moral behavior is susceptible to peer influence. How does information from peers influence moral preferences? We used drift-diffusion modeling to show that peer influence changes the value of moral behavior by prioritizing the choice attributes that align with peers' goals. Study 1 ( N = 100; preregistered) showed that participants accurately inferred the goals of prosocial and antisocial peers when observing their moral decisions. In Study 2 ( N = 68), participants made moral decisions before and after observing the decisions of a prosocial or antisocial peer. Peer observation caused participants' own preferences to resemble those of their peers. This peer influence effect on value computation manifested as an increased weight on choice attributes promoting the peers' goals that occurred independently from peer influence on initial choice bias. Participants' self-reported awareness of influence tracked more closely with computational measures of prosocial than antisocial influence. Our findings have implications for bolstering and blocking the effects of prosocial and antisocial influence on moral behavior.

How conformity can lead to polarised social behaviour

“…Such a model, however, requires either a prohibitive number of trials per participant, or the integration of some other type of information. This problem could possibly be overcome by integrating temporal information to simple choice data: several studies have successfully analysed subjective choices with this method before using so-called sequential sampling models (SSM, see for instance [ 21 , 89 ]). While this approach would require challenging improvements, such as disentangling variability both within and between trials, it could also promote the analysis of other decision components, such as the trade-off between fidelity with one’s preferences and speed in making a decision.…”

“…This means that we may be missing a connection between how close one’s initial attitude is to the observed agent’s and how much she will conform after learning. Indeed, recent research suggests that similarity with the observed agent influences the effect of conformity [ 21 ]. To solve this problem, in future experiments we propose to dynamically adjust the attitude of the agent depending on participants’ own attitude.…”

“…Insights on the cognitive mechanisms behind anti- and prosocial conformity come from the literature on attitude alignment and preference learning [ 14 , 15 ]. These studies have spanned a variety of domains such as attractiveness ratings [ 16 ], food [ 17 ], risk preferences [ 18 – 20 ], moral behaviour [ 21 ], effort [ 20 ], and inter-temporal decisions [ 20 , 22 – 25 ]. At a brain level, learning about others’ attitudes or preferences appears to alter the value representation of choices [ 16 , 21 , 22 , 24 ] or even reward signals [ 19 ], while not necessarily affecting one’s private preferences [ 18 ].…”

“…These studies have spanned a variety of domains such as attractiveness ratings [ 16 ], food [ 17 ], risk preferences [ 18 – 20 ], moral behaviour [ 21 ], effort [ 20 ], and inter-temporal decisions [ 20 , 22 – 25 ]. At a brain level, learning about others’ attitudes or preferences appears to alter the value representation of choices [ 16 , 21 , 22 , 24 ] or even reward signals [ 19 ], while not necessarily affecting one’s private preferences [ 18 ]. In addition, features such as choice variability [ 23 ] or attitude extremeness [ 25 ] seem to be significant predictors of conformity.…”

Human values: A cognitive perspective

“…Human values can and do vary: throughout history (Elias, 1978;Inglehart, 2008;Morris, 2015;Pinker, 2012), throughout the individual's development (House et al, 2013), across populations, cultures, ecologies, moral communities, and individuals (Gelfand et al, 2011;Hofstede, 2001;Inglehart, 2008;Singer, 2011), and across situations within individuals (Aarøe & Petersen, 2014;DeScioli et al, 2014). This suggests that values are sometimes computed to match not an objective standard (e.g., the protein content of food) but a local, possibly variable social consensus-the values that fellow group members are inferred to hold (Yu et al, 2021; see also Kuran, 1997;Zentall & Galef, 1988;Rendell et al, 2010). However, note that socially contingent values appear to reflect variation in how the open parameters of the valuation architecture are filled in from one individual or community or era to the next, rather than variation in the valuation architecture itself.…”