Fig 1 - uploaded by Shai Danziger
Content may be subject to copyright.
Proportion of rulings in favor of the prisoners by ordinal position. Circled points indicate the first decision in each of the three decision sessions ; tick marks on x axis denote every third case; dotted line denotes food break. Because unequal session lengths resulted in a low number of cases for some of the later ordinal positions, the graph is based on the first 95% of the data from each session.
Source publication
Are judicial rulings based solely on laws and facts? Legal formalism holds that judges apply legal reasons to the facts of a case in a rational, mechanical, and deliberative manner. In contrast, legal realists argue that the rational application of legal reasons does not sufficiently explain the decisions of judges and that psychological, political...
Citations
... A range of academic disciplines across different types of social, political, legal, behavioral, and psychological research have sought a more holistic understanding of the factors and 2 processes affecting human decision-making in different social contexts and environments. In this chapter, we review and discuss research that suggests that the thinking and decision-making of legal actors, including judges, attorneys, jurors, probation and parole officials, and expert witnesses, are commonly influenced by broader sociopolitical contexts, their beliefs and misconceptions about particular types of defendants, cognitive shortcuts such as heuristics and intuitions, and their own psychosocial characteristics and moral values (Danziger et al., 2011;Spamann & Klöhn, 2016). Recent advances in computer science and research into the ethics of artificial intelligence in criminal justice are also reviewed, as these developments have begun to address and suggest solutions for the computational shortcomings often seen in algorithmic decision-making and risk assessment tools sometimes used in the criminal justice system (Kleinberg et al., 2017(Kleinberg et al., , 2018. ...
... Today, many U.S. judges at various court levels can consider a wide range of aggravating and mitigating factors that allow them to weigh individual circumstances when deciding sentencing outcomes in particular cases. However, this and other types of sentencing discretion can also result in biases that may seep into judgments and lead to inconsistencies in the sentencing of similarly situated defendants between different judges and jurisdictions (Danziger et al., 2011;Thomaidou & Berryessa, 2023). ...
... Socio-demographic and other characteristics of judges can also meaningfully affect sentencing. For example, according to studies of case law, judges appear to be more punitive when they are tired or hungry (Danziger et al., 2011), when their favorite sports team has lost a game in the days prior to deliberations (Eren & Mocan, 2018), or even when outdoor temperatures are high (Heyes & Saberian, 2019). ...
This is a forthcoming chapter on criminal sentencing for the second edition of the APA Handbook of Forensic Psychology. The chapter begins by describing the historical context, standards, goals, and significance of criminal sentencing in the United States (U.S.). In an effort to elucidate the key influences to and practices by which courts reach sentencing decisions, we then describe psychological and cognitive-behavioral theories relevant to sentencing decision-making. After describing these theories and principles guiding our understanding of the cognitive and computational shortcuts involved in decision-making, we review relevant sentencing research and case law. Key legal, extralegal, and extraneous factors that can influence sentencing are also described to provide a more comprehensive view of sentencing decision-making in practice. We then discuss the significance of sociocultural identities and systemic inequalities in sentencing related to past and current practices, as well as concerns regarding the future of sentencing decision-making that may increasingly rely on automation. Finally, policy issues are discussed, with a particular focus on de-biasing humans, machines, and improving overall sentencing decision-making.
... However, other sources of biases can certainly be more persistent. Birkelund (2014) demonstrates inconsistent evaluations before and after lunch, similar to results found in judges' decisions (Danziger, Levav, and Avnaim-Pesso 2011). In another study, Klein (2003) shows that uninterrupted examination of a large number of tests over an extended period induces grading inconsistency. ...
This study explores the use of artificial intelligence (AI) as a complementary tool for grading essay-type questions in higher education, focusing on its consistency with human grading and potential to reduce biases. Using 70 handwritten exams from an introductory sociology course, we evaluated generative pre-trained transformers (GPT) models' performance in transcribing and scoring students' responses. GPT models were tested under various settings for both transcription and grading tasks. Results show high similarity between human and GPT transcriptions, with GPT-4o-mini outperforming GPT-4o in accuracy. For grading, GPT demonstrated strong correlations with the human grader scores, especially when template answers were provided. However, discrepancies remained, highlighting GPT's role as a "second grader" to flag inconsistencies for assessment reviewing rather than fully replace human evaluation. This study contributes to the growing literature on AI in education, demonstrating its potential to enhance fairness and efficiency in grading essay-type questions.
... Noise arises from idiosyncratic or random factors unrelated to the merit of the case. These may regard personal traits, professional experiences, and work conditions, but also accidental factors, such as cloudy weather (Simonsohn 2007), outside temperature (Heyes and Saberian 2019), decision fatigue (Linder et al. 2014;Philpot et al. 2018), one's birthday (Kato, Jena, and Tsugawa 2020), time passed after a food break (Danziger, Levav, and Avnaim-Pesso 2011), and one's football team losing a match (Chen and Loecher 2019; Eren and Mocan 2018). Decision-makers are affected by these and other factors extraneous to the merit of the decision and hence will rarely repeat the exact same decision twice or make a judgement identical to what another decision-maker would do in the same case. ...
... Machine learning algorithms are widely used in various domains, including entertainment, shopping, healthcare, finance, education, law enforcement, and high-stakes areas like loans [1] and hiring decisions [2,3]. They provide advantages such as tireless performance and the ability to process numerous factors [4,5]. However, algorithms can also exhibit biases, leading to unfair outcomes [6,7]. ...
Machine learning (ML) algorithms have become integral to decision making in various domains, including healthcare, finance, education, and law enforcement. However, concerns about fairness and bias in these systems pose significant ethical and social challenges. This study investigates the fairness of ML models applied to structured datasets in classification tasks, highlighting the potential for biased predictions to perpetuate systemic inequalities. A publicly available dataset from Kaggle was selected for analysis, offering a realistic scenario for evaluating fairness in machine learning workflows. To assess and mitigate biases, three prominent fairness libraries; Fairlearn by Microsoft, AIF360 by IBM, and the What If Tool by Google were employed. These libraries provide robust frameworks for analyzing fairness, offering tools to evaluate metrics, visualize results, and implement bias mitigation strategies. The research aims to assess the extent of bias in the ML models, compare the effectiveness of these libraries, and derive actionable insights for practitioners. The findings reveal that each library has unique strengths and limitations in fairness evaluation and mitigation. By systematically comparing their capabilities, this study contributes to the growing field of ML fairness by providing practical guidance for integrating fairness tools into real world applications. These insights are intended to support the development of more equitable machine learning systems.
... Machine learning algorithms are extensively employed across various fields, including entertainment, shopping, healthcare, finance, education, law enforcement, and critical areas such as loan approvals [1] and hiring decisions [2,3]. They offer benefits like consistent performance and the ability to analyze numerous variables [4,5]. However, these systems are not immune to biases, which can lead to unfair outcomes [6,7]. ...
Machine learning (ML) algorithms play a crucial role in decision making across diverse fields such as healthcare, finance, education, and law enforcement. Despite their widespread adoption, these systems raise ethical and social concerns due to potential biases and fairness issues. This study focuses on evaluating and improving the fairness of Computer Vision and Natural Language Processing (NLP) models applied to unstructured datasets, emphasizing how biased predictions can reinforce existing systemic inequalities. A publicly available dataset from Kaggle was utilized to simulate a practical scenario for examining fairness in ML workflows. To address and mitigate biases, the study employed two leading fairness libraries: Fairlearn by Microsoft, and AIF360 by IBM. These tools offer comprehensive frameworks for fairness analysis, including metrics evaluation, result visualization, and bias mitigation techniques. The research aims to measure bias levels in ML models, compare the effectiveness of these fairness libraries, and provide actionable recommendations for practitioners. The results demonstrate that each library possesses distinct strengths and limitations in evaluating and mitigating fairness. By systematically analyzing these tools, the study contributes valuable insights to the growing field of ML fairness, offering practical guidance for integrating fairness solutions into real world applications. This research underscores the importance of building more equitable and responsible machine learning systems.
... Where do the inconsistencies in public punishment preferences come from? Kahneman et al. (2021) point to several causes, including differential anchors, minor situational cues, and other extra-legal factors such as the temperature, 9 time of the day, or even the performance of the local football team (Danziger et al. 2011;Heyes and Saberian 2019). Moreover, factors contemporarily salient at the moment of decision-making may affect the conceptual sampling of relevant considerations, especially when those considerations are heterogeneous or competing (i.e. ...
Are public preferences for the type or amount of punishment stable? Instability over short periods would complicate empirical desert by undercutting the value of public preferences as policy guides. Using longitudinal, cross-national survey data from Central Europe, we examined within-person stability in punishment preferences along several dimensions: type, amount, and rank order. Individual-level instability was common; respondents frequently changed their punishment preferences across waves. In the aggregate, public opinion was more stable. Our findings support the ‘qualified public input’ model of policy making—aggregate preferences should provide loose guidance for policymakers, with individual-level instability suggesting the ‘latitude of acceptance’ or ‘zone of acquiescence’. Better-educated respondents exhibited more preference stability, thus greater weight should be given to informed public opinion.
... Notes were added by participants to provide further information as to whether ideas should be retained for the second round. To counter the effect of voter fatigue [91], participants were randomly assigned to one of three lists of ideas presented in different orders. Ideas were clustered into broad topics (audio, images, reviewing literature, modeling, data, generating text, negative consequences, citizen science, eDNA, remote sensing, robotics, society, and other) by the lead author. ...
... We attempted to counteract the effect of voter fatigue [5] by randomly assigning participants to receive one of three lists of issues in a different order. We converted the scores into ranks (1-92) and considered issues with median ranks of 1-36 against the qualifying criteria. ...
... We examine the extent to which short term fasting impacts cognitive ability measured in a controlled setting. While previous studies have found that going periods without meals may negatively impact non-hunger related decisions such as leniency of court rulings [8], and monetary impatience [9], we find no significant effect of short run fasting on cognitive ability. ...
... Kuhn, et al. [31] (2014) found a similar treatment effect among glucose deprived subjects, while Levy et al. [32] found that hunger is associated with greater financial risk tolerance. Danziger, et al. [8] suggested that judicial rulings can be swayed by extraneous variables that should have no bearing on legal decisions, such as breaks to eat. The authors found that in sessions before meals, the percentage of favorable rulings gradually declines from 65% to near 0%, then abruptly returns to 65% after a meal break. ...
... Participants in the 3-hour fasting condition were asked to refrain from eating for at least three hours prior to their session with the goal of mimicking the state of hunger that individuals often experience in between meals, similar to the judges studied in Danziger, et al. [8]. On average, participants in this group reported fasting for 8.36 hours (s.d. ...
Growing evidence suggests that resource scarcity can severely impede individuals’ cognitive capacity, resulting in sub-optimal decision making. Few experimental studies investigate whether food deprivation as a form of resource scarcity influences decisions in other non-hunger related domains. We examine the effect of short term fasting on cognitive capacity by exogenously manipulating individuals’ fasting time in a laboratory experiment. Participants were randomly assigned to one of three treatments: 1) 3-hour fast; 2) 12-hour fast; and 3) control, in which participants were not required to fast and consumed a protein shake upon arriving to the lab. Following the manipulation, participants completed the Raven’s Progressive Matrices test which measures cognitive function. Although we find null treatment effects on cognitive ability, our results provide evidence that short term fasting does not directly inhibit cognition.
... The decision-maker deliberates under extreme time restrictions and in highstress environments frequently. According to psychological research, people's decisionmaking skills significantly deteriorate when faced with a large number of decisions to make in a short amount of time [7]. Though they are not affected by human variables, AI systems do not display inconsistencies or absorb pressures. ...
The dynamics underpinning the urban landscape change are primarily driven by social, economic, and environmental issues. Owing to the population’s fluctuating needs, a new and dual perspective of urban space emerges. The Artificial Intelligence (AI) of a territory, or the system of technical diligence associated with the anthropocentric world, makes sense in the context of this temporal mismatch between territorial processes and utilitarian apparatus. This creates cerebral connections between several concurrent decision-making systems, leading to numerous perspectives of the same urban environment, often filtered by the people whose interests direct the information flow till the transformability. In contrast to the conventional methodologies of decision analysis, which are employed to facilitate convenient judgments between alternative options, innovative Artificial Intelligence tools are gaining traction as a means of more effectively evaluating and selecting fast-track solutions. The study’s goal is to investigate the cross-functional relationships between Artificial Intelligence (AI) and current decision-making support systems, which are increasingly being used to interpret urban growth and development from a multi-dimensional perspective, such as a multi-criteria one. Individuals in charge of administering and governing a territory will gain from artificial intelligence techniques because they will be able to test resilience and responsibility in decision-making circumstances while also responding fast and spontaneously to community requirements. The study evaluates current grading techniques and recommends areas for future upgrades via the lens of the potentials afforded by AI technology to the establishment of digitization pathways for technological advancements in the urban valuation.
