Month: August 2022

Upcoming RMME/STAT Colloquium (9/9): Kosuke Imai, “Experimental Evaluation of Algorithm-Assisted Human Decision-Making: Application to Pretrial Public Safety Assessment”

Posted on by

RMME/STAT Joint Colloquium

Experimental Evaluation of Algorithm-Assisted Human Decision-Making: Application to Pretrial Public Safety Assessment

Dr. Kosuke Imai
Harvard University

Friday, September 9, at 11:00AM ET

https://uconn-cmr.webex.com/uconn-cmr/j.php?MTID=m486f7b13e6881ba895b350f338b0c90d

Despite an increasing reliance on fully-automated algorithmic decision-making in our day-to-day lives, human beings still make highly consequential decisions. As frequently seen in business, healthcare, and public policy, recommendations produced by algorithms are provided to human decision-makers to guide their decisions. While there exists a fast-growing literature evaluating the bias and fairness of such algorithmic recommendations, an overlooked question is whether they help humans make better decisions. We develop a general statistical methodology for experimentally evaluating the causal impacts of algorithmic recommendations on human decisions. We also show how to examine whether algorithmic recommendations improve the fairness of human decisions and derive the optimal decision rules under various settings. We apply the proposed methodology to preliminary data from the first-ever randomized controlled trial that evaluates the pretrial Public Safety Assessment (PSA) in the criminal justice system. A goal of the PSA is to help judges decide which arrested individuals should be released. On the basis of the preliminary data available, we find that providing the PSA to the judge has little overall impact on the judge’s decisions and subsequent arrestee behavior. Our analysis, however, yields some potentially suggestive evidence that the PSA may help avoid unnecessarily harsh decisions for female arrestees regardless of their risk levels while it encourages the judge to make stricter decisions for male arrestees who are deemed to be risky. In terms of fairness, the PSA appears to increase an existing gender difference while having little effect on any racial differences in judges’ decisions. Finally, we find that the PSA’s recommendations might be unnecessarily severe unless the cost of a new crime is sufficiently high.

 

Loader Loading...
EAD Logo Taking too long?

Reload Reload document
| Open Open in new tab

RMME Commmunity Members Present New STATA Package: mlmeval

Posted on by

Dr. Anthony J. Gambino (RMME alumnus), Dr. Sarah D. Newton (RMME alumna), and Dr. D. Betsy McCoach (current RMME faculty member) unveiled their new STATA package, mlmeval, at the STATA Conference in Washington, DC this week. Their work pushes the field forward by offering a new tool that provides users with information about both model fit and adequacy for multilevel model evaluation.

 

Abstract:

Model evaluation is an unavoidable facet of multilevel modeling (MLM). Current guidance encourages researchers to focus on two overarching model-selection factors: model fit and model adequacy (McCoach et al. 2022). Researchers routinely use information criteria to select from a set of competing models and assess the relative fit of each candidate model to their data. However, researchers must also consider the ability of their models and their various constituent parts to explain variance in the outcomes of interest (i.e., model adequacy). Prior methods for assessing model adequacy in MLM are limited. Therefore, Rights and Sterba (2019) proposed a new framework for decomposing variance in MLM to estimate R2 measures. Yet there is no Stata package that implements this framework. Thus, we propose a new Stata package that computes both (1) a variety of model fit criteria and (2) the model adequacy measures described by Rights and Sterba to facilitate multilevel model selection for Stata users. The goal of this package is to provide researchers with an easy way to utilize a variety of complementary methods to evaluate their multilevel models.

RMME Community Members Publish Article: Omitted Response Patterns

Posted on by

Merve Sarac (an RMME alumna) and Dr. Eric Loken (a current RMME faculty member) recently published a new article, entitled: “Examining Patterns of Omitted Responses in a Large-scale English Language Proficiency Test” in the International Journal of Testing. Congratulations to Merve and Eric on this excellent accomplishment!

 

Abstract:

This study is an exploratory analysis of examinee behavior in a large-scale language proficiency test. Despite a number-right scoring system with no penalty for guessing, we found that 16% of examinees omitted at least one answer and that women were more likely than men to omit answers. Item-response theory analyses treating the omitted responses as missing rather than wrong showed that examinees had underperformed by skipping the answers, with a greater underperformance among more able participants. An analysis of omitted answer patterns showed that reading passage items were most likely to be omitted, and that native language-translation items were least likely to be omitted. We hypothesized that since reading passage items were most tempting to skip, then among examinees who did answer every question there might be a tendency to guess at these items. Using cluster analyses, we found that underperformance on the reading items was more likely than underperformance on the non-reading passage items. In large-scale operational tests, examinees must know the optimal strategy for taking the test. Test developers must also understand how examinee behavior might impact the validity of score interpretations.