Learning the Preferences of Ignorant, Inconsistent Agents

Authors: Owain Evans, Andreas Stuhlmueller, Noah Goodman

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We present a behavioral experiment in which human subjects perform preference inference given the same observations of choices as our model. Results show that human subjects (like our model) explain choices in terms of systematic deviations from optimal behavior and suggest that they take such deviations into account when inferring preferences.
Researcher Affiliation Academia Owain Evans University of Oxford Andreas Stuhlm uller Stanford University Noah D. Goodman Stanford University
Pseudocode Yes Figure 2: We specify agents decision-making processes as probabilistic programs. This makes it easy to encode arbitrary biases and decision-making constraints. When automated inference procedures invert such programs to infer utilities from choices, these constraints are automatically taken into account. Note the mutual recursion between agent and exp Utility: the agent s reasoning about future expected utility includes a (potentially biased) model of its own decision-making.
Open Source Code No The paper states 'We implemented the model described above in the probabilistic programming language Web PPL (Goodman and Stuhlm uller 2014).' however, it does not provide an explicit statement or link for the open-source code of their specific model implementation.
Open Datasets No The paper describes conducting a 'behavioral experiment in which human subjects perform preference inference' but does not specify a publicly available or open dataset, nor does it provide access details for the data collected from the human subjects.
Dataset Splits No The paper does not provide specific details regarding training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific details about the hardware used to run its experiments.
Software Dependencies No The paper mentions using 'Web PPL (Goodman and Stuhlm uller 2014)' but does not provide a version number for it or for any other key software components.
Experiment Setup No The paper mentions 'For the Bayesian inference corresponding to Equation 1 we use a discrete grid approximation for the continuous variables (i.e. for U, p(s), k and α) and perform exact inference using enumeration with dynamic programming.' however, it does not provide specific hyperparameter values or detailed system-level training settings.