Exercises / Tasks Ideas:
Task 1.Creating a political AI Mirror: The AI decides the nature of the 2nd exercise. Build a “Digital Twin”, with a custom chatGPT chat bot
Build a “Digital Twin” with a custom chatGPT chatbot. A digital twin is a model of an entity in the real world that simulates the behaviour and properties of its physical counterpart. The idea has typically been used to model the behaviour of complex machines or systems, to predict problems or to perform experiments without risk of damage to the real-world counterpart.
This idea has recently gained attention in the area of human judgment: making algorithmic representations of subjective choices reduces noise in a decision-making process. Noise, in this regard, is simply a variation in results when the same decision is repeated. Numerous studies on criminal sentencing by judges show that factors like the time of day, whether a judge is hungry or even how their football team performed on the weekend have vastly different outcomes for the accused. An algorithmic representation of the same judge removes this noise but does not necessarily change the biases influencing the outcome.
Your own Political Digital Twin might give a less noisy representation of your political and moral views and open new possibilities for civic engagement. It’s not practical to poll a population for every decision that is made in a democracy, but what if your digital twin could have a say in the most granular of choices independent of your mood and other random factors?
Steps:
Download the example from github
Update the chatGPT key file.
Modify the parameters to reflect your own views.
Start conversing with the twin and modify the parameters to improve the results.
Task 2 Developing a democratic discourse with AI.
Within your group, think up a scenario or situation where you can incorporate your digital twin. Attempt to play out the situation and document the results. This could be a daily situation where the views and wishes of several people are taken into account, like the organisation and use of the atelier. Were the decisions consistent with your views? Did you think about your own views differently after the experiment: did you discover inconsistencies of biases of your own that you were not aware of?