In a working paper published one year ago1, John J. Horton (MIT) argues that, because of how they are trained and designed, newly-developed large language models (LLM) are implicit computational models of humans, a homo silicus.
LLMs can be used like economists use homo economicus: they can be given endowments, information, preferences, and so on, and then their behavior can be explored in scenarios via simulation.
In another working paper2 shared by Tyler Cowen yesterday, authors take the baton and further elaborate that LLMs should have a deep understanding of fictional characters based on the original literary texts and subsequent commentary.
They prompted OpenAI’s newest large language model, GPT-4, to simulate 148 of those literary fictional characters, starting with William Shakespeare’s Macbeth (1606) and ending with Gary Shteyngart’s Eunice Park (2010), to participate in the Dictator game.
Some “charming” results:
Interestingly, and charmingly GPT-4 understands human characters (at least their literary versions) of any century to be much less selfish than humans.
Literary characters appear to have become less selfish over-time. The Shakespearean characters of the 17th century make markedly more selfish decisions than those of Dickens, Dostoevsky, Hemingway and Joyce, who in turn are more selfish than those of Ishiguro and Ferrante in the 21st.
Male characters were more selfish than female characters: 35% of male decisions were selfish compared to just 24% for female characters. The skew was highest in the 17th century where selfish decisions for male and female were 62% and 20% respectively.
Although I am fundamentally optimistic about the potential applications of LLMs in the future, for the time being their best performance is clearly around the blurred boundaty between reality and fiction.
You can read the comments in Tyler Cowen’s post to understand what I mean…
____________________
(1) Horton, John J. ‘Large Language Models as Simulated Economic Agents: What Can We Learn from Homo Silicus?’ Working Paper. Working Paper Series. National Bureau of Economic Research, April 2023. https://doi.org/10.3386/w31122.
(2) Behavioral Economics and GPT-4: From William Shakespeare to Elena Ferrante #2 (via Marginal Revolution)
Featured Image: NightCafe
Mind the Post 