1. As a first step, we studied how humans teach a very simple abstraction: motifs. Motifs are recurring patterns that can be composed into larger works, like stitch patterns in knitting or clave rhythms in salsa music.

2. Motifs pose a challenge to existing theories of teaching. In classic teaching games, teachers pick out a *part* of a concept to teach the *whole* - like pointing to opposite corners of a rectangle to show a learner where to find it on a canvas.

3. Motifs pose the opposite problem! Imagine you're a knitter, and you show a beginner the rib stitch pattern on a sleeve cuff. The goal isn't to get learners to tell sweaters apart from other things, but to teach a reusable part that can be used to make socks, hat bands, etc.

4. In two experiments (N=446), we studied how people teach and learn motifs in a necklace-making task. Participants created a sample necklace to teach a learner the motifs specific to one community (Exp 1), or saw a sample necklace and tried to infer the underlying motifs (Exp 2)

5. Standard Bayesian models of teaching capture an important aspect of how people teach motifs: They prefer simpler examples than what we would expect from strong sampling. There's no complexity term built in: this preference stems from general communicative principles.

6. Humans also prefer to teach, and learn best from, simple examples. But humans learned best from other humans, compared to examples sampled from pedagogical & strong sampling models, so our theories are still missing part of what makes human teaching so effective!

7. This preprint is our first step towards understanding how teaching empowers learners to solve problems, create, and innovate. We’d appreciate any feedback on this preprint!

@Huang_Ham @BonanZhao @cocosci_lab @natvelali Congrats Ham!🎉

@Huang_Ham @BonanZhao @cocosci_lab @natvelali Congrats Ham! I'm so excited to see the results of all your hard work 🥳