Working Paper: CEPR ID: DP15880
Authors: John Asker; Chaim Fershtman; Ariel Pakes
Abstract: The behavior of artificial intelligences algorithms (AIAs) is shapedby how they learn about their environment. We compare the pricesgenerated by AIAs that use different learning protocols when thereis market interaction. Asynchronous learning occurs when the AIAonly learns about the return from the action it took. Synchronouslearning occurs when the AIA conducts counterfactuals to learn aboutthe returns it would have earned had it taken an alternative action.The two lead to markedly different market prices. When future profitsare not given positive weight by the AIA, synchronous updating leadsto competitive pricing, while asynchronous can lead to pricing closeto monopoly levels. We investigate how this result varies when eithercounterfactuals can only be calculated imperfectly and/or when theAIA places a weight on future profits.
Keywords: No keywords provided
JEL Codes: D43; D82; D83; C72; C73; L13; L41; K21; O33
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| asynchronous learning (C91) | higher prices (D49) |
| synchronous learning (C92) | competitive pricing (L11) |
| asynchronous learning (C91) | supracompetitive pricing (L11) |
| synchronous learning (C92) | Nash pricing levels (D49) |
| differences in pricing outcomes (L11) | intrinsic to updating mechanisms (O30) |
| asynchronous updating (C69) | stable but elevated prices (P22) |
| synchronous updating (C69) | competitive outcomes (L13) |