Experimental Design
Our experiment implements a series of decision problems in networks of bilateral conflicts. Each connection represents a bilateral conflict between two rivals, and a connection weight represents the value of winning that conflict. To win a conflict, agents simultaneously allocate a scarce resource (i.e., budget of 100 tokens) across their rivals. The value of the remaining budget, which is not allocated, is normalized to zero. The winner of each conflict is determined by a lottery contest success function, where an additional unit of resources allocated to a given conflict increases the likelihood of wining that conflict.
In every decision problem, each subject is matched to another three players at random. The experiment consists of 4 blocks of 18 decision problems each. Within a block, subjects face different decision problems each one associated to a different position in various networks. In particular, we consider six non-isomorphic strongly connected network structures that can be built with a set of four nodes (i.e., complete, diagonal, kite, ring, line, and star networks). Across blocks, the only change is the value of the connection between nodes 2 and 3 while all other values remains the same (300 points). The value of the connection changing across blocks is either 300, 600, 900, or 1200 points. And the order in which subjects face these values is randomized for every session.