Testing face-to-face social influence at the limits

Yavuz, Mustafa (2026): Testing face-to-face social influence at the limits. Dissertation, LMU München: Graduate School of Systemic Neurosciences (GSN)

Abstract

Human decisions rarely occur in isolation. In this thesis, I ask when minimal social presence and explicit interaction alter perceptual and value-based decisions, and through which mechanisms. Across three projects, I move from co-visibility, in which another person sees what I see, to face-to-face deliberation, combining behavior, EEG, pupillometry, and speech-text analyses to answer when social context leaves covert signatures in internal computations and when it improves overt performance. In Project 1, I tried to isolate the case of co-perception in a visual decision-making task by contrasting public (co-visible) and private trials without evaluation, feedback, or dialogue. While behavior and drift–diffusion parameters were unchanged across contexts, covert signatures differed: a pre-stimulus neural code for social context emerged, and stimulus-locked pupil dilation was larger on public trials. Together, these findings support an anticipatory social-vigilance account, an expectancy of social relevance under neutral co-visibility, without changes to decision policy. In Project 2, I turned to a difficult-to-communicate sensory modality, namely olfaction, to ask whether collective benefit is achievable in such a domain as well. Using the Sniffin’ Sticks test, I first recorded individual baselines (Day 1) on olfactory discrimination and identification, then had pairs return weeks later to complete the battery again: first privately, then jointly. Dyadic performance exceeded the average of the two individuals, indicating integration beyond simple voting. Collective benefit was positively associated with partners’ ability similarity. Brief dialogue analyses showed that participants spoke more when they disagreed and that, particularly in identification, more concise exchanges were linked to more successful conflict resolution in this lexically constrained modality. Finally, in Project 3, I investigated value-based decision-making under uncertainty using a two-armed bandit paradigm to compare individual versus dyadic decisions and assess whether discussion changes exploration–exploitation policy and overall performance. Results showed that dyads achieved better performance than the average of the individuals. Furthermore, analyses of exploration behavior indicated that the collective advantage was associated with reduced random (undirected) exploration while preserving directed, strategically useful sampling, i.e., discussion stabilized choices without eliminating informative exploration. Taken together, across three projects, I outline a modest but consistent picture, along with clear boundary conditions, of how social context shapes human perception and decision-making. In the co-perception case, without evaluation, feedback, or interaction, I find shifts in preparatory neural signals and arousal states without changes to the decisions. In olfaction, I show that small but reliable collective benefits are achievable with verbal communication, and that these benefits increase with partners’ ability similarity. In value-based choice, discussing options improves accuracy mainly by reducing random exploration while preserving strategically useful sampling. These conclusions are deliberately bounded to dyads, controlled laboratory tasks, and specific manipulations. Within those limits, I specify when merely sharing a scene recalibrates internal readiness and when communication is required to turn private estimates into better joint decisions.

• Item Type: Theses (Dissertation, LMU Munich)
• Keywords: social influence, perception, decision-making
• Faculties: Graduate School of Systemic Neurosciences (GSN)
• Language: English
• Date of oral examination: 9. June 2026
• 1. Referee: Deroy, Ophelia
• Signature of the printed copy: 0001/UMC 32099