Table 3.

Generalized linear model for the probability to change the action in both treatments (shown in figure 5) $(\text{glm}(\text{formula}=\text{change}\sim \text{diff}+\text{info}+\text{rnd}+\text{diff}\times \text{info},\text{family}=\text{binomial}(\text{‘logit}{\hspace{0.17em}}^{\mathrm{\prime }}\text{)}))$. (The significant variables are (i) ‘diff’ revealing that the probability of changing actions is correlated to the difference in pay-offs with the neighbours, and (ii) the combination ‘diff:info’, telling us that this correlation is different in treatments with (TWI) and without (TWO) information.)

variable	estimate	s.e.	z-value	$\text{Pr}(|z|)$
(intercept)	−0.879433	0.050016	−17.583	<2 × 10−16***
diff	0.026458	0.004220	6.269	3.63 × 10−10***
info	−0.012179	0.057056	−0.213	0.831
rnd	0.043624	0.056190	0.776	0.438
diff : info	0.037928	0.006013	6.308	2.83 × 10−10***