Human Decision Making with Machine Assistance: An Experiment on Bailing and Jailing

Grgić-Hlača, Nina; Engel, Christoph; Gummadi, Krishna P.

doi:10.2139/ssrn.3465622

Cited by 43 publications

(50 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The defendant profiles were selected from the COM-PAS dataset, which contained information of 7,214 criminal defendants in Broward County, Florida, USA, between 2013 and 2014 [4,44]. This dataset was widely used by researchers to understand how people interact with machine assistance in their decision making [23,31]. • Forest cover prediction: In this task, participants were shown a geological profile of a wilderness area (in a 30m × 30m cell) containing 8 features-the area's elevation, aspect, slope, hillshade index, the horizontal/vertical distance to nearest surface water, the horizontal distance to nearest roadway, and the horizontal distance to nearest wildfire ignition points.…”

Section: Decision Making Tasksmentioning

confidence: 99%

Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making

Wang

Yin

2021

26th International Conference on Intelligent User Interfaces

155

View full text Add to dashboard Cite

This paper contributes to the growing literature in empirical evaluation of explainable AI (XAI) methods by presenting a comparison on the effects of a set of established XAI methods in AI-assisted decision making. Specifically, based on our review of previous literature, we highlight three desirable properties that ideal AI explanations should satisfy-improve people's understanding of the AI model, help people recognize the model uncertainty, and support people's calibrated trust in the model. Through randomized controlled experiments, we evaluate whether four types of common model-agnostic explainable AI methods satisfy these properties on two types of decision making tasks where people perceive themselves as having different levels of domain expertise in (i.e., recidivism prediction and forest cover prediction). Our results show that the effects of AI explanations are largely different on decision making tasks where people have varying levels of domain expertise in, and many AI explanations do not satisfy any of the desirable properties for tasks that people have little domain expertise in. Further, for decision making tasks that people are more knowledgeable, feature contribution explanation is shown to satisfy more desiderata of AI explanations, while the explanation that is considered to resemble how human explain decisions (i.e., counterfactual explanation) does not seem to improve calibrated trust. We conclude by discussing the implications of our study for improving the design of XAI methods to better support human decision making. CCS CONCEPTS• Human-centered computing → Empirical studies in HCI; • Computing methodologies → Machine learning.

show abstract

Section: Decision Making Tasksmentioning

confidence: 99%

Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making

Wang

Yin

2021

26th International Conference on Intelligent User Interfaces

155

View full text Add to dashboard Cite

show abstract

“…This body of research has used interviews, surveys, and experiments to empirically probe people's perceptions towards algorithmic decisions. Examples include studies on how humans perceive algorithmic decisions versus human decisions in managerial contexts [34], whether people perceive certain features (such as criminal history or neighborhood safety) as fair to be used to predict criminal risk [25,50], how explanation styles might matter in shaping people's justice perceptions [9], how members of traditionally marginalized communities feel about algorithm (un)fairness [54], how affected communities feel about algorithmic decisions in the context of a child welfare system [12], how the general public perceives online behavioral advertising that used demographic factors (e.g., race) as targeting variables [44], which statistical definitions of fairness people perceive to be the fairest in the context of loan decisions [47], as well as how humans use AI systems to make decisions [23,24]. This past body of work mostly used storyboards or text to present several algorithmic scenarios to their study participants, often without tackling the results and performance of the underlying machine learning models.…”

Section: Human Perceptions Towards Algorithmic Decisionsmentioning

confidence: 99%

Designing Alternative Representations of Confusion Matrices to Support Non-Expert Public Understanding of Algorithm Performance

Shen

Jin

Cabrera

et al. 2020

Proc. ACM Hum.-Comput. Interact.

View full text Add to dashboard Cite

Ensuring effective public understanding of algorithmic decisions that are powered by machine learning techniques has become an urgent task with the increasing deployment of AI systems into our society. In this work, we present a concrete step toward this goal by redesigning confusion matrices for binary classification to support non-experts in understanding the performance of machine learning models. Through interviews (n=7) and a survey (n=102), we mapped out two major sets of challenges lay people have in understanding standard confusion matrices: the general terminologies and the matrix design. We further identified three sub-challenges regarding the matrix design, namely, confusion about the direction of reading the data, layered relations and quantities involved. We then conducted an online experiment with 483 participants to evaluate how effective a series of alternative representations target each of those challenges in the context of an algorithm for making recidivism predictions. We developed three levels of questions to evaluate users' objective understanding. We assessed the effectiveness of our alternatives for accuracy in answering those questions, completion time, and subjective understanding. Our results suggest that (1) only by contextualizing terminologies can we significantly improve users' understanding and (2) flow charts, which help point out the direction of reading the data, were most useful in improving objective understanding. Our findings set the stage for developing more intuitive and generally understandable representations of the performance of machine learning models. CCS Concepts: • Human-centered computing → Human computer interaction (HCI).

show abstract

“…The authors concluded that laypersons performed no worse than COMPAS. Successive studies replicated this finding [40,43,53]. However, they also found that participants' predictive performance was considerably lower than the RAI's when outcome feedback was removed, the base rate was decreased, or when the area under the curve (AUC), instead of accuracy, was considered [43].…”

Section: Introductionmentioning

confidence: 96%

“…In one of these studies [38], the defendant's description and the RAI's recommendation were presented to participants at the same time. In another [40], the RAI's information was made available only after the participant had made an initial prediction. However, the psychology and behavioral economics literature suggest that participants might rely on the RAI more heavily in the former setting due to a phenomenon known as the anchoring effect, a.k.a.…”

Section: Introductionmentioning

confidence: 99%

“…Problematically, for these tasks, even participants that answer carefully can achieve low predictive accuracy and thus accuracy may not be a reliable proxy for effort. Alternatively, one might consider the time spent by participants on the entire survey, which many past studies have reported [38,40,53,81]. By taking into account the number of defendants shown in each of these surveys, a quick computation suggests that, on average, the prediction for a single defendant may take less than 15 seconds.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The Impact of Algorithmic Risk Assessments on Human Predictions and its Analysis via Crowdsourcing Studies

Fogliato

Chouldechova

2021

Proc. ACM Hum.-Comput. Interact.

View full text Add to dashboard Cite

As algorithmic risk assessment instruments (RAIs) are increasingly adopted to assist decision makers, their predictive performance and potential to promote inequity have come under scrutiny. However, while most studies examine these tools in isolation, researchers have come to recognize that assessing their impact requires understanding the behavior of their human interactants. In this paper, building off of several recent crowdsourcing works focused on criminal justice, we conduct a vignette study in which laypersons are tasked with predicting future re-arrests. Our key findings are as follows: (1) Participants often predict that an offender will be rearrested even when they deem the likelihood of re-arrest to be well below 50%; (2) Participants do not anchor on the RAI's predictions; (3) The time spent on the survey varies widely across participants and most cases are assessed in less than 10 seconds; (4) Judicial decisions, unlike participants' predictions, depend in part on factors that are orthogonal to the likelihood of re-arrest. These results highlight the influence of several crucial but often overlooked design decisions and concerns around generalizability when constructing crowdsourcing studies to analyze the impacts of RAIs.

show abstract

Human Decision Making with Machine Assistance: An Experiment on Bailing and Jailing

Cited by 43 publications

References 0 publications

Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making

Are Explanations Helpful? A Comparative Study of the Effects of Explanations in AI-Assisted Decision-Making

Designing Alternative Representations of Confusion Matrices to Support Non-Expert Public Understanding of Algorithm Performance

The Impact of Algorithmic Risk Assessments on Human Predictions and its Analysis via Crowdsourcing Studies

Contact Info

Product

Resources

About