Bounded Rationality: What generative AI is really doing for you

One of the most important principles in AI and, arguably, in daily life is the principle of Bounded Rationality. The term was coined by Herb Simon when discussing models of perfectly rational humans in economics in the late 50s. He defined it as follows:

Broadly stated, the task is to replace the global rationality of economic man with the kind of rational behavior that is compatible with the access to information and the computational capacities that are actually possessed by organisms, including man, in the kinds of environments in which such organisms exist.

Herb Simon, 1957

If we drill into the “computational capacities” part of the statement, though, there is a simple concept that often escapes people:

You do not know all the consequences of the things you know.

We generally assume we know what things can/cannot happen or that make sense in a given situation. On closer examination, this turns out to almost never be true.

Let’s imagine you know all the facts about something, and you also know most of the rules that would tell you how those facts can be combined. An example would be that “A is a liquid” and “B is a sponge” + that “Sponges normally absorb liquid.” Even with a relatively small set of facts and a small set of “rules”, it very quickly becomes impossible to imagine all possible consequences and combinations of the facts.

The problem we face in any work, knowledge, or creative task is “How to explore the space of possibilities?”. As humans, we absolutely forced to use smart guesses, direct our attention and leave out many combinations. We simply don’t think of them because we don’t have time. Our biases also mean that there are many combinations we will not even imagine.

This is what the current wave of generative AI is doing:

1. Pulling in new facts (sources) relevant to a particular question. Like relevant images you didn’t know of or facts/text that you had never read.
2. Combining things in ways that your brain had not come up with yet. This is really applying rules to combine things that you had forgotten about or didn’t know.

In other words, generative AI is allowing us to jump around the space of possibilities in a different way to just sitting in a chair and thinking things through. This is an amazing superpower. Some of the results can be inspiring and enable us to create things we would not have thought of.

Of course, each generation of technology will hit its own limits. AI is also bounded in its rationality. We’ve only just figured out the most basic ways for generative AI to stitch together facts. In many cases, results can still be highly non-sensical.

So think of generative AI as an accelerated and tunable exploration device rather than a second brain.