Benchmarks like the ones proposed in the Stanford paper could help teams better judge fairness in AI models—but actually fixing those models could take some other techniques. One may be to ...