Automatic Philosophising

Automatic Philosophising

The following philosophical musings were generated by a machine learning language model called GPT-2. They were created by a very weak version of GPT-2 which was released to the public in May 2019. The full version has nearly 5 times as many nodes and produces much more coherent text (sometimes, at least!) As philosophers, there’s always the spectre of the hoax paper – whether from an academic seeking to expose areas of research which can be perceived as engaging in obfuscation to conceal a vacuous centre, or from students who don’t have the time or inclination to compose an essay. Markers and reviewers be warned.

I asked GPT-2 for some views on rationality and human nature, and it turned to entirely an fictionalised version of David Hume for answers. Everything that is presented here was written entirely by GPT-2. I have removed all prompt sections. No edits were made within paragraphs.

Humans are highly intelligent. We have evolved to make rational decisions. And yet many of us succumb to the urge to act upon irrational impulses. In his famous essay on “Why We are Here,” David Hume wrote that when the world is “socially and psychologically balanced, all that remains is to choose from a variety of modes of reasoning to guide our actions. This may sometimes be accomplished by trial and error, by careful reasoning and reflection on the grounds of experience; or, it may be accomplished by means of a series of verbal or written suggestions which serve but to confirm or reinforce our preconceptions.”1 It looks like the same could be said of our capacity to make irrational choices. There is already growing evidence of how deeply ingrained irrationality is.


But even when the decision to make a choice is informed by sound reasoning, our rationality is not always rewarded. In a 2013 study , researchers from Oxford University tracked the behaviour of more than 80 adults across a 12-week period. Using functional magnetic resonance imaging to measure both brain activity in the pre- and post-test rounds of choice making, the researchers then compared the participants’ activity over that same time period to an array of brain scans to gauge the extent to which the choices affected their performance, even if the participants had not actually made themselves a choice at that point.


After the initial scan, the participants’ brains were scanned again, and the scans showed that the activity of their frontal lobes – the part of the brain used for decision-making – was decreased in all participants at the final scan. It’s the frontal lobes’ role in helping us make rational choices that time of choice, suggesting that the decision-making process involved in making a choice had ceased. “We believe this could be a mechanism for why participants are less motivated in later months after a decision,” researchers wrote. When those frontal lobes are working for a certain task, they’re more likely to continue working in the face of new stimuli. “If you give people an opportunity to make one choice in the future, but then you tell them that that decision won’t be made at all – it ends in failure,” says Renshaw. “That’s what’s going on in the brain at this point. You are having to make multiple decisions, each against your judgment, until you’re forced to try one more. That’s when the prefrontal activity is least likely to be functioning.”


I’m not sure how far science and medicine can go in explaining why our decisions are influenced by emotion rather than reason, but I do know that we can find a good place to begin.