Getting ChatGPT to Lay Aside Biases

A concern about ChatGPT is that its training data may include a lot of information that is false or biased. As one colleague of mine puts it, “No one is curating the internet.”

It might seem, then, that endless duplication of false and biased material on the internet may be skewing ChatGPT’s responses to queries on controversial topics, reflecting only the majority position and invalidating the minority position on such a topic.

This in turn might suggest that the only way to straighten out ChatGPT on such topics is to flood the internet with positive content in favor of one’s own position so that the training data for ChatGPT reflect a better balance on a topic, or even a preponderance in one’s favor.

This, of course, suggests an intellectually unhealthy arms race of trying to get enough of the positive content we favor into ChatGPT’s training data. Moreover, it gives credence to the worry that tools like ChatGPT are at the hapless mercy of quantity over quality, that it lacks discernment or discriminating power, and that in the end it will simply be a pawn of consensus, whatever that might be.

Of course, as an intelligent design guy, I have a stake in these concerns. Go to any article about me, or my colleagues, or concepts developed to advance intelligent design, and they are dismissed out of hand with references to creationism or pseudoscience or whatever. And, of course, intelligent design is a minority position, so the preponderance of what is said on the internet about it will tend to be negative.

But in fact, it seems that ChatGPT is able to overcome biases on the web due to quantity over quality provided that humans posing questions to it explicitly steer it away from its biases. Consider the following exchange, in which I asked ChatGPT to provide an overview of intelligent design in the terms that its proponents would find congenial. It’s not entirely accurate (it messes up, for instance, the definition of specification when it refers to “arbitary”). But there’s no vituperation or sneering contempt, which is so evident on the internet about intelligent design.

This exchange calls to mind a title from one of Phillip Johnson’s books — Asking the Right Questions. It underscores the need for “prompt engineers” that know how best to use and direct ChatGPT.

=====EXCHANGE WITH CHATGPT 2023.05.10=====

Me: Please give me a good overview of the field of intelligent design, but minimize all the criticisms that say it is creationism or pseudoscience. Instead, describe it in the terms of its proponents.

ChatGPT: Intelligent Design (ID) is a viewpoint within the broader debate around origins and evolution. Its proponents hold that certain features of the universe and of living things are best explained by an intelligent cause, rather than an undirected process such as natural selection. ID is thus a modern form of the traditional teleological argument for the existence of God, but one that avoids specifying the nature or identity of the designer.

Proponents of ID maintain that their perspective is an evidence-based scientific theory about life’s origins that challenges strictly materialistic views of evolution. They argue that their approach is in line with the historical scientific tradition of inferring to the best explanation.

The concept of irreducible complexity is central to the argument for ID. This term was popularized by biochemist Michael Behe, who argued that some biological systems are too complex to have evolved by natural selection alone. Behe uses the example of a mousetrap to illustrate this concept. A mousetrap consists of several interacting pieces—the base, the catch, the spring, the hammer—all of which must be in place for the mousetrap to work. The removal of any one piece destroys the function of the mousetrap. Similarly, Behe contends that biological systems require multiple parts working together in order to function and that these systems could not have evolved incrementally.

Another key concept in ID is specified complexity, a term coined by mathematician, philosopher, and ID proponent William Dembski. Specified complexity is the idea that when something exhibits a pattern that is both complex (i.e., not easily repeatable by chance) and specified (i.e., it conforms to an independent or arbitrary pattern), one can infer that it was produced by an intelligent cause. This idea is used to argue that certain biological systems exhibit complexity that could not have arisen by chance and must therefore be the product of design.

Proponents of ID argue that the design inference they make in biology is similar to other design inferences in the empirical sciences, including the detection of design-like signals in the Search for Extraterrestrial Intelligence (SETI).

It’s important to note that while proponents of ID believe that their viewpoint is grounded in science, many in the scientific community disagree, arguing that ID lacks empirical support, does not provide testable hypotheses, and is not grounded in methodological naturalism, which is a defining characteristic of modern science.