There are two coding projects that I’ve had on my mind for the last little while, bothering me like a loose tooth. Here they are:
- New strategies for the iterated prisoner’s dilemma in Netlogo. NetLogo allows you to model agent based interactions following the incentives set up in the PD scenario. The publicly available scenario has a limited number of strategies: defector, tit-for-tat, and so on. The strategies have limited memory capacity, so, e.g., you can have a tit-for-tat strategy, who responds to anyone who defected on them in the past by defecting. I’ve added a few more strategies.
- What I need now is to add in two new strategies that are responsive to incentives: the ‘desperado’ and the ‘turncoat’. The idea is that the desperado cooperates unless they achieved a below-average payoff on the previous round, in which case they defect; and the turncoat does the opposite, defecting if they achieved an above-average payoff the previous round.
- Revisions to explanatory coherence. Thagard has created a model of explanatory coherence called ECHO, which is part of a set called COHERE and coded in LISP. The model is meant to capture the process of coherentist inference in scientific reasoning, with a unique accent on outputting theories that fit a certain threshold (approximating a particular vision of inference to the best explanation). There are several axioms that the model is meant to operationalize, among them the principles of symmetry, explanation, and analogy. They are as follows: (1) Symmetry: for any two propositions, the coherence of P1 with P2 implies the coherence of P2 with P1 – there is no such thing as a one-way coherence relationship. (2) Explanation: if a set of premises purportedly explain some conclusion, then the premises must cohere with the conclusion, and with each other. However (assuming these premises are co-hypotheses), as the number of premises goes up, the degree of coherence goes down – simplicity is better than complexity. (3) Analogy: for any two separate premise-conclusion pairs (P1-C1 and P2-C2), if the premises are alike (P1&P2), and the conclusions are alike (C1&C2), then they cohere.
- The first problem I have with it is that it doesn’t run on modern Mac OS. I would like to have a version that can work.
- More importantly, I’d like to have a version of ECHO that allows you to dial down the weight of these three principles. In other words, I want to see what theories pass the threshold for best explanation for reasoners who deny one or more of the three factors mentioned above (symmetry, explanation, and analogy).