When Does Uncertainty Matter?: Understanding the Impact of Predictive Uncertainty in ML Assisted Decision Making
Abstract: As machine learning (ML) models are increasingly being employed to assist human decision makers, it becomes critical to provide these decision makers with relevant inputs which can help them decide if and how to incorporate model predictions into their decision making. For instance, communicating the uncertainty associated with model predictions could potentially be helpful in this regard. In this work, we carry out user studies (1,330 responses from 190 participants) to systematically assess how people with differing levels of expertise respond to different types of predictive uncertainty (i.e., posterior predictive distributions with different shapes and variances) in the context of ML assisted decision making for predicting apartment rental prices. We found that showing posterior predictive distributions led to smaller disagreements with the ML model's predictions, regardless of the shapes and variances of the posterior predictive distributions we considered, and that these effects may be sensitive to expertise in both ML and the domain. This suggests that posterior predictive distributions can potentially serve as useful decision aids which should be used with caution and take into account the type of distribution and the expertise of the human.
License: Creative Commons Attribution 4.0 International (CC BY 4.0)
Submission Length: Long submission (more than 12 pages of main content)
Changes Since Last Submission: **Changes for the camera ready version (02 Jun 2023)** * Added a paragraph in Section 5.1 (second last paragraph) to address the action editor’s suggestions * Replaced “City A” with “Cambridge, Massachusetts” throughout the text and in Figure 3 * Added an Acknowledgements section * Removed the blue font coloring (which tracked the revisions) * Removed duplicate entries in the reference list * Fixed formatting errors in the references * Fixed several typos, grammatical mistakes, and inconsistencies in language usage throughout the paper **Changes for the first revision (23 Apr 2023)** All changes are in blue text. The main changes are: * We re-launched our user study and now have 1,330 responses from 190 participants, twice that of our original study. * We updated the claims made throughout paper to reflect the insights obtained from the larger study we conducted. * We added a thorough discussion of limitations of our work in the Discussion. Other changes: * Reduced the number of ML expertise categories from 3 to 2 * Clarified our motivation of ML assisted decision making under limited information * Clarified that participants were not aware of what features the ML model uses to predict apartment prices * Clarified why participants were not exposed to multiple experimental conditions * Added details on the model accuracy in the training examples * Clarified that participants were not told the true rental prices in the testing examples * Clarified that the relative size of the apartments is also given to the participants * Added the median time to complete the study * Clarified that participant ID was the only random effect included in our regression analyses * Reduced the number of self-reported trust and usefulness categories from 5 to 3 * Fixed typos and grammatical mistakes
Assigned Action Editor: ~Laurent_Charlin1
Submission Number: 845