| Popis: |
This experiment was designed to study the extent to which people use social learning (i.e., the incorporation of information/knowledge of others in their network) to solve complex problems. The experiment uses a multi-armed bandit problem, translated into a game in which participants have to learn which of four card decks has the most profitable cards. Participants get 20 tries to draw a card from 1 of the 4 card decks, so learning is possible through repetition: with each card drawn, the subject increases insights into the type of cards in the deck and the deck's profitability. Social learning is introduced by connecting each participant to three others, from whom they could see, each round, which deck they chose and what card they had drawn. This increased the amount of information from which to learn which deck is the most profitable one. Subjects were divided in network positions that varied in the amount of local clustering and betweenness centrality to whether bridging tie positions (higher likelihood of new information) increased the chances of learning which card deck was most profitable. The experiment tested two treatments: a competitive and a noncompetitive one. In the competitive one, the participants would also see how their total payoff ranked in comparison to that of all (9) other participants, and would be told that the three best performing players would receive a bonus in the end, while for the three worst performing players points would be subtracted. These treatments were used to test if participants were more likely to rely on information from their network in competitive contexts. In total 200 subjects participated in 10 different sessions conducted in the ELSE lab of Utrecht University, between Feb. 9 and 23, 2016. All subjects played both treatments. The data was used for an article on the effect of bridging tie positions that was published in Social Networks: Vriens, E. & Corten, R. (2018). Are bridging ties really advantageous? An experimental test of their advantage in a competitive social learning context. Social Networks, 54: 91-100. doi.org/10.1016/j.socnet.2018.01.007 The archive contains the original ztree scripts, the raw data, the stata dataset, and stata and mplus scripts of the analyses reported in the Social Networks paper. |
