Skip to content

PTT-Calculator v1.0

Latest
Compare
Choose a tag to compare
@klee-schoeppl klee-schoeppl released this 23 Feb 10:01
· 1 commit to main since this release
ffba8d6

Automated analysis of probabilistic truth-table tasks for studies on reasoning under uncertainty. Written as research assistant on BMBF project <01UL1906X> of Dr. Dr. Niki Pfeifer. This first version (Feb 2022) of the PTT-Calculator offers analysis for up to 10-case setups.


This current version can be accessed for testing in the browser here, hosted by shinyapps.io. If you want to actually work with the software, please download it and run it locally.

This small program is meant to assist in creating studies in which participants are asked to evaluate the probability point-values or uncertainty intervals of sentences featuring two propositions (the antecedent A and the consequent C) connected by various sentence-connectives. Participants are shown a sentence of the form 'A [sentence connective] C', as well as all the possible outcome-cases ('A ∧ C', 'A ∧ ¬C', '¬A ∧ C' or'¬A ∧ ¬C'), some of which remain hidden ('?'). They are then asked to determine with which probability (intervals) they assume the sentence to hold.

The PTT-Calculator aims to automatically calculate the uncertainty intervals or probability point-values of such sentences for a variety (of interpretations) of sentence connectives, which I attempted to group according to their natural language counterparts. In addition, for each setup the program also calculates the values of 10 different measures of argument strength (or factual support) between the antecedent and the consequent. Keep in mind that not all of these probabilistic measures share the same value range.


How to use?

First, select the type of sentence connective you are interested in, or the placeholder '[any connective]' in case you are interested in all interpretations. Next, select a total number of cases, and choose how many of them make both the antecedent and consequent true ('A ∧ C'), only one of the two ('A ∧ ¬C' or '¬A ∧ C'), neither ('¬A ∧ ¬C') and how many of them are hidden ('?'). Finally, decide whether to include the calculation of halfway interpretations (see Pfeifer 2013b, Pfeifer forthcoming) and notions of argument strength (see Pfeifer 2013a, Sprenger & Hartmann 2010). Press 'Calculate Task' to create and view tables containing the desired data. For easy of use, the program can also display the formulae used for each output, and offers each output-table as pre-generated LaTeX code.

How it works:

Calculation of task setups featuring no uncertainty - that is, no [?] cases - is pretty straight-forward. Here, one can simply determine the probabilities required (e.g., P(A), P(¬C), P(A ∧ C) and so on) by counting up the relevant cases and dividing them by the total number of cases in the task. For tasks featuring uncertainty though, the application opens a grid - akin to a propositional logic truth table - containing every possible variation behind the hidden sides. As each [?] side essentially functions as two boolean variables (A or ¬A and C or ¬C), this amounts to a 2n times 2^(2n) grid, where n equals the number of [?] sides in given task. Reading out this grid then allows calculation of the probabilities required (e.g., P(A), P(¬C), P(A ∧ C) and so on) for each of the 2^(2n) no-uncertainty-setups compatible with the task at hand. The application then calculates the interpretations and notions of factual support for each of these possibilities individually, in order to return the correct minimal, maximal, median and mean values.

For more detailed explanations and application examples, see Pfeifer 2013b and Pfeifer & Tulkki 2017.


References

  • Hartmann, S., & Sprenger, J. (2010). Bayesian epistemology.
  • Howson, C., & Urbach, P. (1993). Scientific Reasoning: the Bayesian Approach, Open Court. Lasalle, IL.
  • Pfeifer, N. (2013a). On argument strength. In F. Zenker (Ed.), Bayesian argumentation. The practical side of probability (p. 185-193). Dordrecht: Synthese Library Vol. 362 (Springer).
  • Pfeifer, N. (2013b). The new psychology of reasoning: A mental probability logical perspective. Thinking & Reasoning, 19(3-4), 329-345.
  • Pfeifer, N. & Tulkki, L. (2017). Abductive, causal, and counterfactual conditionals under incomplete probabilistic knowledge. In Gunzelmann, G., Howes, A., Tenbrink, T., &, Davelaar, E. (Eds.). Proceedings of the 39th Cognitive Science Society Meeting (p. 2888-2893).