Abstract
Individuals can rely on social learning to improve problem solving performance. Individuals' opportunity to engage in social learning is constrained by their communication networks which are shown to shape the efficiency of the problem solving process. To this date, experimental and simulation-based research disagrees on what kind of network structure is to be preferred, both providing support for efficient network structures that allow immediate diffusion of good solutions as well as for inefficient networks that prevent premature diffusion of solutions that turn out to be poor (Lazer and Friedman 2007; Mason and Watts 2012; Barkoczi and Galesic 2016). These existing studies, however, implicitly assume that agents indiscriminately copy when observing others’ superior solutions, an assumption that is not empirically grounded (Acerbi et al. 2016; Derex et al. 2016). We propose a simple derivation of existing simulation frameworks by introducing a known human bias, an 'Ikea-effect' (Norton et al. 2012), such that agents prioritize individual learning over social learning when solving problems. Our simulations show that the biased Ikea agents tend to outperform their rational competitors in both efficient and inefficient networks and that this more realistic search mechanism thus explains discrepancies in previous results. Overall, our study illustrates how one can optimize collective problem solving not only by manipulating the structural level, but also by embodying an empirically grounded learning bias at the individual level.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have