Do we have free will or are our behaviors predetermined? Are we as free as we think we are?

These are the questions that can be asked when we talk about the Aaronson’s oracle, an apparently simple algorithm that, in spite of limiting itself to studying which keys we press, is capable of knowing which ones we are going to press next.

It may seem simple and uninteresting, but considering that a simple computer program is capable of knowing how we are going to behave based on how we are responding, it is no small matter. Let’s look at it below.

What is the Aaronson oracle?

Aaronson’s oracle consists of a computer program which has proven to have a high capacity for predicting human decisions .

The algorithm behind this program was developed by Scott Aaronson and, by means of a task to be done by the participant, the program is able to know which key is going to be pressed next. The person is in front of a computer with the program on and must press the keys D or F as many times as he wants and in the order he wants .

While the person is pressing keys, the oracle will give feedback, indicating whether the key pressed was the one he had in mind or not. In other words, the oracle indicates whether you were right in predicting that the person would press the D key or the F key.

How does it work?

As we have seen, despite the mystery of the name, Aaronson’s oracle is nothing more than an algorithm behind a computer program. This program is responsible for analyzing the 32 different possible sequences of five letters, formed by the keys D and F , that the person has typed before. The algorithm memorizes them as the subject types them and, when the person types again a sequence that begins in a similar way to one already done, the algorithm predicts the next letter.

To understand it better, let’s put the following case. We have at some point typed in the following sequence D-D-D-F-F. The algorithm will have memorized it and, if we have just typed the following sequence D-D-D-F-F, the oracle will most probably establish that the next key pressed will be another F. Of course, we could type D and make the oracle make a mistake, but it can be said that, advanced the sequences, the percentage of prediction of the algorithm is higher than 60% .

When we press the first keys, the oracle’s prediction percentage will not be high. This is due to the fact that we have just put information, that is, there are no previous sequences and, therefore, there is no background that can be linked to the information immediately put. On the first attempt, it is impossible for the oracle to predict whether we are going to put a D or an F. This decision can be totally random, and therefore the oracle will not have a certainty of more than 50%.

However, once we have put in several key sequences, the program will predict our behaviour pattern more accurately . The more keys pressed, the more information and therefore the more able we are to know whether the next thing will be a D or an F. In its web version you can see the success percentages. If they are less than 50%, it means that the oracle is not correct, and if they are higher, it means that it is on the right track.

The surprising thing about the program is that, although we can try to make it confusing, the algorithm learns from it . It ends up using our decision against us, making us see that, although we had supposedly done it freely, this is not really the case.

Are we that predictable?

Based on what we have seen with Aaronson’s oracle, which consists of a simple computer algorithm, it is necessary to open the debate on whether the human being, who has always shown his free will, really has such a gift or, on the contrary, is just an illusion.

The idea behind the concept of free will is that people behave totally independently of their previous acts and stimuli present in their immediate and closest environment. That is, regardless of what we have done or what we see, hear or feel, our behaviors can be consciously decided and detached from the past and the environment . In short, free will means that nothing is written, that everything is possible.

The opposite of this concept is the idea of determinism. What we have done before, what we have already experienced or what we are experiencing right now determines our actions. However conscious and masterful we may think we are of our behaviors, according to determinism, they are nothing more than the result of what has already happened. They are the next link in a chain of events that are each the cause of the next.

Looking at these definitions, one may think that yes, indeed, the idea that yesterday, last week, every day of the previous month or even for years we have eaten at two o’clock in the afternoon is a fact that will most probably be repeated tomorrow, but this does not mean that it will determine what happens tomorrow. That is to say, although it is very likely that we will eat at two o’clock tomorrow, it does not mean that we cannot change, in a totally random way, the time at which we will eat the following day.

However, what Aaronson’s oracle brings to light is that human beings, although we try not to be predictable, end up being so . Even trying to prevent a simple computer program from knowing which key we are going to press, just by pressing the other one, we are already being predictable, since the computer has beaten us to it. We have already given it enough information to know how we are going to behave.

Antegrade Amnesia and Repeated Behavior: The Case of Mary Sue

Some time ago, a woman became famous for, unfortunately, a symptom of her transitory global amnesia that turned out to arouse the curiosity of the network. The lady, named Mary Sue, appeared in a video recorded by her daughter, in which she was having a conversation.

So far so normal, except for one important detail: the conversation was repeated in a loop, and lasted about nine and a half hours . Mary Sue repeated herself like an old cassette tape. Luckily for the woman her amnesia was solved after one day.

This type of repeated conversations are common in people who suffer from anterograde amnesia and, in fact, have been widely documented, besides serving to shed some light on the problem that concerns us here: are our decisions free? The problem that prevents us from checking whether a decision we have made in the past was the result of our supposed free will or, on the contrary, was determined, is that we cannot travel to the past and try to change it.

But, fortunately, cases like Mary Sue’s allow us to understand this a little better. Mary Sue was, metaphorically speaking, in a time loop. She would talk, time would pass, and suddenly it was like she was going back in time. Back at the beginning, Mary Sue started asking the same questions, saying the same answers . Suffering from anterograde amnesia, she could not generate new memories, so her brain was constantly resetting itself and, having the same triggering events, it carried out the same behaviour.

With Mary Sue’s case we could come to the conclusion that we are not free, that the idea of free will is nothing more than an illusion and that it is completely normal that algorithms like Aaronson’s Oracle, and any others that are being manufactured, are capable of knowing how we are going to behave.

This same issue has been addressed in a more scientific way in the outstanding work of Koenig-Robert and Pearson (2019). In their experiment they managed to predict the decisions of the experimental subjects up to 11 seconds in advance , but not before the behaviour itself, but before they were even aware of their own choice.

Nevertheless, and by way of a final reflection, it is important to say that, although interesting, no computer program or experiment will be able to resolve, in a categorical way, a philosophical debate as old as the world itself. Although scientific research has helped to understand the human being, it is really difficult to understand how we come to behave in natural situations, and not in laboratory contexts.

Scott Aaronson and Computer Science

Scott Joel Aaronson is a computer scientist and professor at the University of Texas at Austin. His area of research is primarily quantum computing. He has worked at MIT and has conducted postdoctoral studies at the Institute for Advanced Study and the University of Waterloo, USA.

He has won several awards for his research, receiving the Alan T. Waterman Award in 2012, as well as the Award for the best Scientific Article on Computing in Russia in 2011, for his work The Equivalence of Sampling and Searching . Among his most outstanding works is the Complexity Zoo, a wiki that catalogues various calculations belonging to the theory of computational complexity .

He is the author of the blog Shtetl-Optimized , as well as having written the essay Who Can Name the Bigger Number ? (“Who Can Name the Bigger Number?”), a paper which has been widely disseminated in the world of computer science, and uses the concept of the Beaver Algorithm, described by Tibor Radó, to explain the limits of computability using more pedagogical language.

Bibliographic references:

  • Koenig-Robert, R., Pearson, J (2019). Decoding the contents and strength of imagery before volitional engagement. Sci Rep 9, 3504 https://doi.org/10.1038/s41598-019-39813-y
  • Aaronson, Scott. (2014) “Who Can Name the Bigger Number?”. academic personal website. Electrical Engineering and Computer Science, MIT.