Learning When to Advise Human Decision Makers

Authors: Gali Noti, Yiling Chen

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The results of a large-scale experiment show that our advising approach manages to provide advice at times of need and to significantly improve human decision making compared to fixed, non-interactive, advising approaches.
Researcher Affiliation Academia Gali Noti1,2 , Yiling Chen1 1Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University 2The School of Computer Science and Engineering, The Hebrew University of Jerusalem {galinoti, yiling}@seas.harvard.edu
Pseudocode No No pseudocode or clearly labeled algorithm block found in the paper.
Open Source Code No The data are available on the authors websites.
Open Datasets Yes As the risk-assessment component of the algorithmic assistant, we use the model of [Green and Chen, 2019b] that was trained on 47,141 defendant cases from a dataset collected by the U.S. Department of Justice [DOJ., 2014] (top diagram in Figure 1a)... For learning the advising policy, we train a randomforest model on experimental data of human predictions from [Green and Chen, 2019a].
Dataset Splits No For learning the advising policy, we train a randomforest model on experimental data of human predictions from [Green and Chen, 2019a]... The training data consist of 6,250 predictions made by 250 human participants, for 500 defendant cases.
Hardware Specification No No specific hardware details (GPU/CPU models, memory, or detailed computer specifications) are mentioned in the paper.
Software Dependencies No No specific software dependencies with version numbers are mentioned in the paper (e.g., Python 3.8, PyTorch 1.9, CPLEX 12.4).
Experiment Setup No The paper describes the experimental treatments and participant numbers but does not provide specific experimental setup details like hyperparameter values, model initialization, or specific optimizer settings.