with Lara Suraci, Robin Cubitt, and Chris Starmer.
- In a lottery-choice environment (think of robo-advisors), right from the start, most people are happy to delegate to an expected-value-maximising algorithm.
- Clearly less people delegate to an algorithm that „relies on modern decision science“ to try and choose what their previous choices suggest they would choose (while avoiding clearly suboptimal choices)
- People stick to their choices of delegating or not over rounds; they have a higher tendency to switch away from an algorithm after a zero-payoff, but only in case of the preferential algorithm
- We can exclude dissatisfaction with the algorithms‘ choices as a main driver of the treatment effects
