Axel, Ockenfels ; Christoph, Engel ; Max, ; Christoph, Engel ; Max R. P., Grossmann ; Axel, Ockenfels. (2023) Integrating machine behavior into human subject experiments: A user-friendly toolkit and illustrations.
In: Discussion Paper Series of the Max Planck Institute for Research on Collective Goods. RePEc:mpg:wpaper:2024_01.

Full description at Econpapers

To set an attribute on all Threads of a Conversation, we use convo.all.var, not merely convo.var. Thus, we only provide a composite-like interface, not a true composite (Gamma et al., 1994: sec. 4.3). We distinguish between attributes set on the Conversation and its Threads because Python does not expose a complete Proxy-like mechanism as does JavaScript, and Threads indeed each obtain their own reference to the underlying object, or a copy if the object is sufficiently simple. For Table 3, we used Polars (www.pola.rs) to create a DataFrame from the returned list of dicts, after which we grouped by politician and time and summarized to obtain GPT-4’s mean belief about the public approval rating. All data is returned from the shorthand as a “list of dicts” to be processed by other software or exported to any spreadsheet format.