“Causal reasoning is critical to machine learning,” said Nailong Zhang, a software engineer at Meta. Meta uses causal inference in a machine learning model that governs how many and what types of notifications Instagram should send to its users to keep them coming back.
Romila Pradhan, a data scientist at Purdue University in Indiana, uses counterfactuals to make automated decision-making more transparent. Organizations now use machine learning models to choose who gets loans, jobs, parole, even housing (and who doesn’t). Regulators have begun requiring organizations to explain the results of many of these decisions to those affected by them. But reconstructing the steps taken by a complex algorithm is difficult.
Pradhan believes counterarguments can help. Let’s say a bank’s machine learning model rejects your loan application, and you want to know why. One way to answer that question is through counterarguments. Given that an application was denied in the real world, would it have been denied in a fictional world where your credit history was different? What if you had a different zip code, job, income, etc.? Building the ability to answer such questions into future loan approval programs, Pradhan says, will enable banks to offer customers reasons beyond just yes or no.
Counterfactuals are important because it’s how people think about different outcomes, Pradhan says. “They’re a good way to get explanations.”
They can also help companies predict people’s behavior. Because counterfactuals make it possible to infer what might happen in a given situation, not just the average, technology platforms can use it to pigeonhole people more accurately than ever before.
The same logic that can separate the impact of dirty water or lending decisions can be used to improve the impact of Spotify playlists, Instagram notifications, and ad targeting. If we play this song, will that user listen longer? If we show this picture, will that person continue to spin? “Companies want to understand how to make recommendations to specific users, not just the average user,” says Gilligan-Lee.