When in doubt follow the crowd? A mouse-tracker study to uncover the role of uncertainty in Social Influence in risk estimations.

In the age of social media, peer reactions are instant and unambiguous “ this post is good. This is bad”. This looming instant judgement can fuel feelings of uncertainty about what to do. In this study we want to test whether the amount of felt uncertainty influences susceptibility to (negative) peer influence.

Adolescents and young adults are prone to engage in risk behaviours more than any other age groups. Studies have shown that the influence of peers is strong when it comes to estimating the risks, and subsequently engaging in the risk behaviours. In fact, influence from peers is cited as one of the strongest predictors of maladaptive behaviours (i.e., excessive alcohol and drug use (Colby et al, 2005), skipping school (Xu & Fan, 2018), internet bullying (Williams & Guerra, 2007), unhealthy eating (Wouters, Larsen, Kremers, Dagnelie, & Geenen, 2010), and smoking (Alexander, Piazza, Mekos, & Valente, 2001)). Interventions to reduce these undesirable behaviours have thus far yielded small and mixed results (e.g., Colby et al., 2005). In order to combat such behaviours we need a better insight into what makes young adults and adolescents most vulnerable to social influence.

Decades of social influence research suggested that conforming to others´ judgment and norms is primarily motivated by the desire to belong to the modelled group (i.e., the normative influence account), and by the desire to reduce uncertainty about what the appropriate behaviour is (i.e., the informative influence account; e.g., Deutsch & Gerard, 1955; Cruwys, Bevelander & Hermans, 2015). Especially when it comes to risk estimations, that lack an objective standards, people tend to look for others as a source of information. However, how uncertain do we need to be before we adopt other peoples´ opinions, conform to their norms, and act in ways that might affect our well-being?

This knowledge gap in psychological research is partially due to the difficulty of measuring uncertainty (e.g., Ciranka & van den Bos, 2019). Explicit questions might encounter reluctance to report it, or are hindered by a lack of insight in one’s own uncertainty. New technologies, such as mouse-tracker software provide fresh opportunities to unobtrusively measure if uncertainty is indeed causing susceptibility to social influence.

In our study we can directly measure the magnitude of uncertainty with the mouse-tracker while young adults make risk estimations of behaviours. We will use this measure to predict whether a) they will change their estimation in the face of a majority of deviant peers, and b) whether their uncertainty goes down when they see that others agree with them.

Research methodology Participants between 18 and 24 years of age are invited to estimate the riskiness of 16 behaviours, once privately and once while they are exposed to peer information. These behaviours have been extensively pilot tested to elicit uncertainty about their riskiness (e.g., Knoll et al., 2015; Knoll et al., 2017). The specific peer information is connected to their first estimations, such that participants in the social influence condition will see incongruent peer information for half of the trials, and congruent peer information for the other half. Participants in the control condition will simply rate the same 16 behaviours again. The peer information is presented in a bargraph that indicates what the majority (between 68 and 86%) of previous participants had selected. This study has received ethical approval from the Aarhus University ethics committee. A detailed analysis plan can be found in the pre-registration form https:// aspredicted.org/blind.php?x=p22kr9

Sample size An a priori power analysis indicated a minimum sample size of 132 to find a medium effect size (with .80 power, ɑ = .05) for conformity (between-subjects). Multiple power analyses for the different analyses were run and our sample size was based on the largest sample size. The aim is to collect 10-20% more to account for any participants dropping out or technical errors, leading to a total sample size of 150 participants.

Expected costs . £262.50 incl. vat.
Dissemination The data, analysis scripts and programming scripts will be available through a folder on OSF. We collect no personal information, so this is not in conflict with GDPR rules.