Working Paper: NBER ID: w30633
Authors: Susan Athey; Dean Karlan; Emil Palikot; Yuan Yuan
Abstract: Online platforms often face the challenge of being both fair (i.e., non-discriminatory) and efficient (i.e., maximizing revenue). Using computer vision algorithms and observational data from a micro-lending marketplace, we find that the choices that online borrowers make when creating online profiles impact both of these objectives. We further support this finding with a web-based randomized survey experiment. In the experiment, we create profile images using Generative Adversarial Networks that differ in a specific feature and estimate the impact of the feature on lender demand. We then evaluate counterfactual platform policies based on the changeable profile features, and identify approaches that can ameliorate the fairness-efficiency tension.
Keywords: online marketplaces; fairness; efficiency; user preferences; profile images
JEL Codes: D0; D40; J0; J02; O1
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| policies encouraging smiling images (J18) | improved fairness (measured by Gini coefficient) (D63) |
| policies promoting certain features (J18) | less equitable outcomes (D63) |
| profile images that include a smile (Y90) | better funding outcomes (I22) |
| bodyshot images (Y90) | worse funding outcomes (I24) |
| style features (smiling vs. bodyshot) (Y90) | funding outcomes (I22) |
| borrower types (e.g., gender, race) (G51) | style features (smiling vs. bodyshot) (Y90) |
| style features (smiling vs. bodyshot) (Y90) | existing inequities in funding outcomes (I24) |