“…However, these results are at odds with the prior research into iterated learning, which has repeatedly shown that learners have a bias for simplicity-not informativenessand that iterated learning therefore gives rise to simple, degenerate, uninformative languages. Informative languages only emerge in the presence of a shared communicative task (e.g., Carr et al, 2017;Kirby et al, 2015;Motamedi et al, 2019;Raviv et al, 2018;Saldana et al, 2019;Winters et al, 2018) or an artificial analog of such a task (Beckner et al, 2017;Kirby et al, 2008), neither of which was present in the two studies described by Carstensen et al (2015). One reason why learners might be biased toward simplicity is that, when learners are faced with understanding the world, the best strategy-given that they have no expectations about how the world is structured-is to apply Occam's razor; all things being equal, simpler hypotheses should be preferred over more complex ones (Li & Vitányi, 2008;Rissanen, 1978;Solomonoff, 1964).…”