Articles tagged with Fairness
A Cornell University-led study using the Tetris game found that players who receive fewer turns perceive their partner as less likable, regardless of whether a human or AI allocates the turns. The researchers also discovered that machine allocation behavior can lead to worse performance and gameplay when fairness is not guaranteed.
Researchers found that machine-learning models trained with descriptive data label rule violations more harshly than humans, leading to potential serious implications in the real world. This study highlights the need for careful consideration of data labeling and training methods to ensure fairness and accuracy in AI decision-making.
A team of experts identified 29 sources of bias in AI/ML models for medical imaging, including data collection, preparation, and deployment. The study provides a comprehensive roadmap for mitigating these biases and ensuring fairness, equity, and trust in AI/ML models.
A new study reveals that first-generation college graduates often originate from more affluent families with greater resources. This challenges the notion of a meritocracy, where success is solely based on individual effort. The research highlights the significant advantage these students have due to their family's socioeconomic status.
A Carnegie Mellon University professor proposes a new AI model that highlights the evidence of fairness in decision-making. The model provides counterfactual explanations to demonstrate whether past decisions were fair or biased. This research has practical implications for legislative efforts, industry norms, and AI model evaluation.
An international research team from the Global South emphasizes the importance of diversity, equity and inclusion in tropical ecology and conservation. They suggest ten actions for researchers living in the Global South to improve their inclusivity and participation in international research collaborations.
A new study analyzes carbon dioxide removal quotas for developing countries, considering issues of equity and fairness. The allocation methods were based on extended equity principles and assessed the implications for these seven tropical developing countries.
Researchers propose an alternative redistricting procedure that takes input from two major political parties and results in quantifiably fairer district maps. The bisection protocol, inspired by the
The special issue explores key characteristics of ethical codes, their underlying values, and challenges in implementing them. Studies highlight common ethical values across cultures, emphasizing the importance of autonomy and care in early childhood education.
A new IIASA-led study estimates that global climate goals can be met in a cost-effective manner with fair financial flows, but current pledged funding is insufficient. The researchers recommend increasing financing from rich countries to poorer regions to meet the Paris Agreement targets under most equity considerations.
Deep learning models can become less accurate in recognizing specific categories of images, sounds, or text after network pruning. Researchers demonstrate a technique to address this challenge, improving the fairness of deep learning models.
A new study by Anglia Ruskin University found that hostile behavior from leaders can lead to co-workers adopting similar behavior, creating a toxic workplace environment. The research also discovered an association between experiencing hostile behavior and emotional exhaustion, job insecurity, and abusive peer behavior.
A study published in Cell Press found that when humans are involved, computer decisions are perceived as fairer. Participants deemed decisions related to positive outcomes fairer than negative ones and had concerns over fairness in systems with higher stakes. The results suggest that automated decision-making systems need careful desig...
A new University of Illinois project aims to improve undergraduate students' ability to estimate their knowledge using artificial intelligence methods. The researchers will utilize machine learning to anticipate student performance and provide personalized feedback to enhance studying strategies.
Research reveals significant differences in earnings between white and ethnic minority workers within the same workplace. The study suggests that changes in pay-setting arrangements are needed to bring parity between workers.
A large-scale study led by the University of Leicester found that people with a strong sense of fairness cheat less, regardless of winning or losing. The researchers examined the behavior of over 500 participants in lab-based games and discovered that only those with low 'inequality aversion' were more likely to cheat.
Researchers at MIT found that explanation methods used to aid human decision-makers in high-stakes situations often have lower accuracy for minoritized subgroups. The fidelity of these explanations varies dramatically between subgroups, with the quality often significantly lower for women and Black people.
New research finds that journalists and laypersons describe racial and gender inequalities in terms of disadvantage, while wealth inequality is described in terms of advantage. The framing of inequalities is linked to their perceived legitimacy, with illegitimate inequalities being framed more strongly as disadvantages.
A major study explores biases against women, minorities, and Indigenous knowledges in Australian history on Wikipedia. The research aims to improve representation and make the platform more equitable.
A new study found that symbolic counting skills are strongly related to young children's fair sharing behavior. Researchers developed a counting intervention that improved sharing behavior in preschoolers, suggesting that learning to count may be a key factor in developing this social skill.
A study by Carnegie Mellon University researchers found that lawyers who view negotiation as a game, with arbitrary rules, are less willing to disclose honest information. Lawyers with higher moral character are more likely to adopt an ethical approach, while those with lower moral character are more prone to game framing.
A University of Arkansas professor is developing learning algorithms for building fair decision models in both offline and online learning settings. He will utilize Pearl's Structural Causal Model to analyze causal effects from observational data and propose universal formulations for measuring long-term fairness.
A simulated crisis-care event revealed the moral distress experienced by triage team members who had to prioritize patients for scarce resources. The study aimed to operationalize a process for making life-and-death patient decisions, but triage-team members struggled with balancing individual patient needs with fair resource allocation.
Researchers at MIT created a process called DualFair that can remove bias from data used to train machine-learning models. The method tackles both label bias and selection bias, significantly reducing discrimination in loan predictions while maintaining high accuracy.
The University of Huddersfield and Vietnam's ULIS are conducting a research collaboration to ensure fairness and inclusivity in English language assessments. The project aims to develop an online toolkit and community of practice for teachers, improving student outcomes across Vietnam.
A new study by the University of Southampton finds that skilled overseas workers with families face a higher risk of poverty than UK workers and their net incomes are lower under the UK's points-based immigration system. The researchers warn that this could make the UK less attractive for overseas workers who could fill skills gaps.
A new study led by Boston Medical Center researchers shows that Black individuals who were incarcerated have a statistically significant 65% increase in mortality rates compared to non-Hispanic non-Black individuals. This finding highlights the long-term impacts of incarceration on health, particularly among marginalized communities.
Researchers developed a fair-diverse allocation optimization framework to optimize the distribution of COVID-19 vaccines, treatments, and testing supplies. The model aims to get limited resources to the most vulnerable subgroups while ensuring fairness and diversity, independent of demographic background.
A new study found that most incarcerated individuals' in-prison experiences lessen their perceptions of procedural justice and fairness, regardless of the length of time served. The research used a nationally representative sample of over 31,000 adults and examined various in-prison experiences on inmates' perceptions of staff members'...
Researchers developed a new statistical engine to evaluate gymnastics judges' accuracy and fairness. The study found that some judges are significantly better than others at judging, with the best-trained judges being only two to three times more accurate than their peers.
A recent study published in Nature Machine Intelligence challenges the long-held assumption that accuracy and fairness are mutually exclusive in machine learning. Researchers found that optimizing models for accuracy does not necessarily compromise fairness, particularly when adjustments are made to data, labels, and scoring systems.
A new study explored how prosecutors think about race in criminal justice, finding that a color-blind approach can entrench racial disparities. The study suggests that changing prosecutor views on race and core beliefs is necessary to achieve justice.
A study by Matthew Auer highlights the need for community-led wildfire protection plans to address growing threats, particularly among low-income households. These plans can provide educational resources, fire-resistant home construction, and access to funding.
A study examining Pennsylvania death penalty cases found that victim's gender and race affected every outcome. Cases with white female, male or black male victims were more likely to receive a death sentence than those with black female victims. The study also found that prosecutors were more likely to seek the death penalty in cases w...
Researchers found that local stakeholders consider fairness in distribution of money from marine protected areas according to who holds rights over the area, rather than equality or costs incurred. This challenges common assumptions in conservation literature and practice, emphasizing the importance of explicit identification of local ...
A recent study by the University of Cambridge found that individuals who are more concerned about contracting COVID-19 tend to be more disapproving of others' wrongdoings. The researchers suggest that concerns about health and safety shape morality, leading to non-rational judgments.
A randomized survey experiment found heightened awareness of immigration detention fosters beliefs that the US immigration system lacks procedural and outcome fairness. Individuals' intentions to migrate were not influenced by heightened awareness of immigration detention or nonjudicial removals.
Researchers discovered that fair treatment alone is insufficient to foster a sense of value in groups; instead, recognizing individual talents and qualities provides a sense of inclusion while promoting mental health. This approach complements the idea of 'fitting in' by allowing individuals to stand out and feel valuable.
Employees perceive prosocially motivated supervisors as fairer and are more likely to give them the benefit of the doubt, even when they make mistakes. Research from the University of Notre Dame found that supervisor motives impact fairness judgments and that focusing on prosocial motives can buffer the negative effects of low justice ...
Achuta Kadambi's article in Science explores how medical device physics can perpetuate bias across racial and gender lines. He suggests quantifying sample fairness and recalibrating performance metrics to address these issues.
Bias in medical devices results in undesirable performance variation across demographic groups, influencing health inequality. Medical engineers can learn from computer science's approach to address bias in devices.
A team led by Michigan State University's Pang-Ning Tan is working to make artificial intelligence algorithms more fair and aware of their potential biases. They're developing a definition of fairness that considers perspectives from groups of people, rather than just individual outcomes.
Experts argue that a 'regulatory ecosystem' is necessary for trustworthy AI, rather than relying on public education or increased transparency. Drs Knowles and Richards propose the use of AI documentation, such as IBM's AI Factsheets, to empower specialists in assessing trustworthiness.
Researchers at NYU are developing new methods and tools to address public sector policy impacts, including assessing and correcting biases in city housing inspections, policing, and courts. The project aims to create open-source tools for eliminating implicit biases and enhancing human decision-making.
A new method developed by Cornell University researchers can inject fairness into political redistricting. The 'Fairmandering' data tool creates billions of potential electoral maps for each state and then algorithmically identifies possibilities meeting desired criteria for fairness.
Researchers and practitioners will gather to discuss foundational challenges in data science, including prediction, inference, fairness, and ethics. The event features keynote speakers on AutoML, NLP, and machine learning in healthcare, as well as papers on algorithmic fairness and ecosystem forecasting.
Researchers from Michigan State University found that people who want to improve their empathy tend to shift towards more liberal ideologies, prioritizing care and fairness. This study is the first to explore shifts in personalities and morals due to volitional change.
A new tool, FairCo, was developed to improve the fairness of online rankings by giving equally relevant choices roughly equal exposure and avoiding preferential treatment for high-ranked items. This can correct unfairness in existing algorithms and curtail personal choice.
Customers with excellent dining experiences express the most dissatisfaction with automatic gratuities. Research found that high-quality service does not compensate for negative customer responses to non-voluntary tipping systems. Restaurants can explore ways to give customers a sense of control, such as providing feedback or adding se...
The report, Digital Political Ethics: Aligning Principles with Practice, recommends best practices for digital campaigns to promote fair elections. It offers guidelines on issues like microtargeting, data transparency, and security of voter data.
A new framework for designing machine learning algorithms helps prevent undesirable behavior, including biases in medical diagnoses and financial predictions. The approach shifts the burden from users to designers, making it easier to create 'well-behaved' algorithms without requiring extensive domain knowledge or complex data analysis.
Researchers at Stanford and UMass Amherst develop a new technique to create machine-learning algorithms that can learn to avoid undesirable outcomes such as gender bias and excessive risk. Their approach, called the Seldonian algorithm, enables users to specify what behaviors they want an AI system to avoid with high probability.
Researchers at UMass Amherst and Stanford have designed a new framework for fairer, safer machine learning algorithms. The Seldonian framework allows users to specify safety and fairness constraints, enabling the development of more responsible AI applications.
Researchers found that hiring algorithms are often opaque and biased, with few vendors disclosing concrete information on validation and mitigation. The study encourages transparency and conversation around ethical decision-making in pre-employment assessments using machine learning.
Children between 3-4 years old value generosity and benevolence as a general norm, while older children (5-6) develop a more complex concept of fairness that includes reciprocity. The study suggests that reciprocity norms are not fully developed until around age 5.
A rigged card game was used to study how people perceive inequality and fairness. The study found that winners were more likely to believe the game's outcome was fair, even when it was heavily tilted in their favor.
Top female athletes have higher testosterone levels, which enhance physical performance and raise questions about fairness in competition. The European Society of Endocrinology recommends a maximum testosterone level of 5.0 nmol/L for female athletic events to ensure competitive fairness.
According to a new study, managers with empathy, integrity, and trustworthiness see increased productivity and positive team dynamics. Servant leaders prioritize staff development, creating loyal and empowered teams.
A recent study published in Nature Communications challenges prior research on moral decision-making by showing that people's moral behavior can shift depending on the context. The researchers identified four moral strategies: inequity aversion, guilt aversion, greed, and moral opportunism, which involves switching between these strate...
A study by Lancaster University and Memorial University found that women who wear the hijab or niqab while testifying in court are perceived as more credible than uncovered women. The researchers suggest several possible explanations for this bias, including the idea that religious garments signal honesty and dispel common rape myths.