Working Paper: NBER ID: w31921
Authors: João Guerreiro; Sérgio Rebelo; Pedro Teles
Abstract: We consider an environment in which there is substantial uncertainty about the potential adverse external effects of AI algorithms. We find that subjecting algorithm implementation to regulatory approval or mandating testing is insufficient to implement the social optimum. When testing costs are low, a combination of mandatory testing for external effects and making developers liable for the adverse external effects of their algorithms comes close to implementing the social optimum even when developers have limited liability.
Keywords: Artificial Intelligence; Regulation; Externalities
JEL Codes: H21; O33
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Regulatory approval (G18) | Developers' incentives to create risky algorithms (D82) |
| Holding developers accountable for adverse external impacts (D62) | Social optimum outcomes (D61) |
| Mandatory beta testing + Developer liability (C88) | Outcomes close to social optimum (D61) |
| Limited liability (K13) | Developers' risk-taking behavior (D81) |