Letters: Is It Ethical to Use Artificial Intelligence to Predict Human Outcomes?

Readers question a machine’s ability to solve man’s problems

Thomas Pajot / Shutterstock / Dan Cretu / Getty / Katie Martin / The Atlantic

Artificial Intelligence Shows Why Atheism Is Unpopular

In July, Sigal Samuel wrote about new artificial-intelligence modeling projects that may help predict policy outcomes—particularly around issues of religious pluralism.

In his book Sapiens, Yuval Noah Harari defines two classes of chaotic systems. In level-one chaotic systems, the rules can be complex, but they operate deterministically and produce a result based on the initial conditions. Level-two chaotic systems react to predictions about themselves and therefore can never be predicted accurately. A recent example might be the 2016 U.S. presidential election, where FBI Director James Comey assumed that the candidate Hillary Clinton was going to win; he claims that he used this prediction to justify the release of information to the public, in order to guard against the losing candidate claiming bias on the part of the FBI.

Politics is a second-order chaotic system, and does not always operate deterministically in the same way that a natural system does, because the actors’ beliefs in the future change the behavior of the system, altering the outcomes.

If a charismatic leader is able to persuade a significant portion of a population that they are able to create a certain outcome in the future, that shared belief can significantly alter the system outcome in unpredictable ways.

Jonathan Gross
Edina, Minn.

I agree with the physicist [Neil Johnson] that the modeler is overstating the model’s power. And regarding the modeler’s struggle with ethics—he’s not treading on any new territory here. Inventions can virtually always be used for good or evil; look no further than Alfred Nobel and his invention of dynamite. Nevertheless, it’s interesting stuff and the modeler is right in that if he doesn’t do it, someone else surely will.

Kirk Lohnes
Bosque Farms, N.M.

I loved the article. As soon as I read, “An international team of computer scientists, philosophers, religion scholars, and others are collaborating to build computer models that they populate with thousands of virtual people, or ‘agents,’” I thought of Isaac Asimov’s award-winning Foundation series of science-fiction novels. I’m surprised it wasn’t mentioned.

The projects described in the article are the beginning of “psychohistory,” in Asimovian terms, including the ethical dilemmas of the researchers and the potential for ill-use.

Eric Lawton
Toronto, Ontario

Samuel uses the word secular without really defining it, especially when she identifies the “four factors” needed to secularize a society. What exactly does that mean? Eliminating religion? The increase of people who identify as “spiritual but not religious”? Or, as Charles Taylor argues, maximizing pluralism? The assumption throughout the article is that being religious is an obstacle to pluralism, but one can be religious and accept pluralism. I think she should be more careful in the use of this term as its highly debated, and, at least, provide a working definition.

Aristotle Papanikolaou
Fordham University
New York, N.Y.