🎯 Top Personalized Recommendations
Harvard University andor
Why we think this paper is great for you:
This paper delves into algorithmic fairness, particularly in the sensitive domain of criminal justice, which is highly relevant to your interest in ethical AI applications. It explores different fairness concepts and their conflicts, offering insights into optimizing for both fairness and accuracy.
This paper delves into algorithmic fairness, particularly in the sensitive domain of criminal justice, which is highly relevant to your interest in ethical AI applications. It explores different fairness concepts and their conflicts, offering insights into optimizing for both fairness and accuracy.
Abstract
Algorithmic fairness has grown rapidly as a research area, yet key concepts remain unsettled, especially in criminal justice. We review group, individual, and process fairness and map the conditions under which they conflict. We then develop a simple modification to standard group fairness. Rather than exact parity across protected groups, we minimize a weighted error loss while keeping differences in false negative rates within a small tolerance. This makes solutions easier to find, can raise predictive accuracy, and surfaces the ethical choice of error costs. We situate this proposal within three classes of critique: biased and incomplete data, latent affirmative action, and the explosion of subgroup constraints. Finally, we offer a practical framework for deployment in public decision systems built on three pillars: need-based decisions, Transparency and accountability, and narrowly tailored definitions and solutions. Together, these elements link technical design to legitimacy and provide actionable guidance for agencies that use risk assessment and related tools.
Algorithmic fairness has grown rapidly as a research area, yet key concepts remain unsettled, especially in criminal justice. We review group, individual, and process fairness and map the conditions under which they conflict. We then develop a simple modification to standard group fairness. Rather than exact parity across protected groups, we minimize a weighted error loss while keeping differences in false negative rates within a small tolerance. This makes solutions easier to find, can raise predictive accuracy, and surfaces the ethical choice of error costs. We situate this proposal within three classes of critique: biased and incomplete data, latent affirmative action, and the explosion of subgroup constraints. Finally, we offer a practical framework for deployment in public decision systems built on three pillars: need-based decisions, Transparency and accountability, and narrowly tailored definitions and solutions. Together, these elements link technical design to legitimacy and provide actionable guidance for agencies that use risk assessment and related tools.
AI Summary
- The paper proposes a modified group fairness objective that minimizes weighted error loss while keeping false negative rate differences within a specified tolerance (τ) across protected groups, aiming to improve solution feasibility and overall predictive accuracy. [2]
- A practical framework for deploying fair public decision systems is introduced, based on three pillars: need-based decisions, transparency and accountability, and narrowly tailored definitions and solutions, linking technical design to legitimacy. [2]
- The paper highlights that individual fairness (similar individuals treated similarly) and group fairness (equal outcomes/error rates across groups) are often in conflict, especially when the statistical distance between feature distributions of different groups is large. [2]
- The proposed modification to group fairness explicitly introduces a tolerance bound (τ) for false negative rate differences, allowing for a trade-off between strict parity and overall accuracy, while also surfacing the ethical choice of error costs (α, β). [2]
- The "Three Pillars" framework emphasizes that fairness is a value-based, context-dependent notion, requiring specific definitions and solutions justified by historical context and transparently communicated to stakeholders. [2]
- Critiques of canonical group fairness definitions include inherent biases in training data leading to feedback loops, the inevitable "latent affirmative action" when addressing different base rates, and the practical infeasibility of achieving fairness across an explosion of demographic subgroups. [2]
- Group Fairness: An algorithm does not treat different demographic groups systematically differently, often defined by equal error rates (e.g., demographic parity, equalized odds, equal opportunity, calibration within groups). [2]
- Individual Fairness: Similar individuals are treated similarly, often formalized via a Lipschitz condition where the statistical distance between outcome distributions for two individuals is bounded by their input feature distance. [2]
- Process Fairness: An algorithm gains legitimacy through an open and transparent design and implementation process, focusing on input and methodology rather than just output. [2]
- Modified Group Fairness Objective: Minimizing a weighted error loss (αFN + βFP) subject to a tolerance bound (τ) on the absolute difference in false negative rates between any two protected groups. [2]
Department of Computing
Why we think this paper is great for you:
You will find this paper highly relevant as it directly investigates the critical trade-off between fairness and accuracy in edge AI environments. It addresses a core challenge in deploying equitable AI systems in real-world, dynamic settings.
You will find this paper highly relevant as it directly investigates the critical trade-off between fairness and accuracy in edge AI environments. It addresses a core challenge in deploying equitable AI systems in real-world, dynamic settings.
Abstract
Federated learning (FL) has emerged as a transformative paradigm for edge intelligence, enabling collaborative model training while preserving data privacy across distributed personal devices. However, the inherent volatility of edge environments, characterized by dynamic resource availability and heterogeneous client capabilities, poses significant challenges for achieving high accuracy and fairness in client participation. This paper investigates the fundamental trade-off between model accuracy and fairness in highly volatile edge environments. This paper provides an extensive empirical evaluation of fairness-based client selection algorithms such as RBFF and RBCSF against random and greedy client selection regarding fairness, model performance, and time, in three benchmarking datasets (CIFAR10, FashionMNIST, and EMNIST). This work aims to shed light on the fairness-performance and fairness-speed trade-offs in a volatile edge environment and explore potential future research opportunities to address existing pitfalls in \textit{fair client selection} strategies in FL. Our results indicate that more equitable client selection algorithms, while providing a marginally better opportunity among clients, can result in slower global training in volatile environments\footnote{The code for our experiments can be found at https://github.com/obaidullahzaland/FairFL_FLTA.
Federated learning (FL) has emerged as a transformative paradigm for edge intelligence, enabling collaborative model training while preserving data privacy across distributed personal devices. However, the inherent volatility of edge environments, characterized by dynamic resource availability and heterogeneous client capabilities, poses significant challenges for achieving high accuracy and fairness in client participation. This paper investigates the fundamental trade-off between model accuracy and fairness in highly volatile edge environments. This paper provides an extensive empirical evaluation of fairness-based client selection algorithms such as RBFF and RBCSF against random and greedy client selection regarding fairness, model performance, and time, in three benchmarking datasets (CIFAR10, FashionMNIST, and EMNIST). This work aims to shed light on the fairness-performance and fairness-speed trade-offs in a volatile edge environment and explore potential future research opportunities to address existing pitfalls in \textit{fair client selection} strategies in FL. Our results indicate that more equitable client selection algorithms, while providing a marginally better opportunity among clients, can result in slower global training in volatile environments\footnote{The code for our experiments can be found at https://github.com/obaidullahzaland/FairFL_FLTA.
Princeton University, RTX
Why we think this paper is great for you:
This paper is a strong match for you as it critically examines biases and errors in LLMs, especially concerning demographic factors and medical applications. It highlights the challenges in ensuring generalizable and ethical AI behavior.
This paper is a strong match for you as it critically examines biases and errors in LLMs, especially concerning demographic factors and medical applications. It highlights the challenges in ensuring generalizable and ethical AI behavior.
Abstract
Recent research has shown that hallucinations, omissions, and biases are prevalent in everyday use-cases of LLMs. However, chatbots used in medical contexts must provide consistent advice in situations where non-medical factors are involved, such as when demographic information is present. In order to understand the conditions under which medical chatbots fail to perform as expected, we develop an infrastructure that 1) automatically generates queries to probe LLMs and 2) evaluates answers to these queries using multiple LLM-as-a-judge setups and prompts. For 1), our prompt creation pipeline samples the space of patient demographics, histories, disorders, and writing styles to create realistic questions that we subsequently use to prompt LLMs. In 2), our evaluation pipeline provides hallucination and omission detection using LLM-as-a-judge as well as agentic workflows, in addition to LLM-as-a-judge treatment category detectors. As a baseline study, we perform two case studies on inter-LLM agreement and the impact of varying the answering and evaluation LLMs. We find that LLM annotators exhibit low agreement scores (average Cohen's Kappa $\kappa=0.118$), and only specific (answering, evaluation) LLM pairs yield statistically significant differences across writing styles, genders, and races. We recommend that studies using LLM evaluation use multiple LLMs as evaluators in order to avoid arriving at statistically significant but non-generalizable results, particularly in the absence of ground-truth data. We also suggest publishing inter-LLM agreement metrics for transparency. Our code and dataset are available here: https://github.com/BBN-E/medic-neurips-2025-demo.
Recent research has shown that hallucinations, omissions, and biases are prevalent in everyday use-cases of LLMs. However, chatbots used in medical contexts must provide consistent advice in situations where non-medical factors are involved, such as when demographic information is present. In order to understand the conditions under which medical chatbots fail to perform as expected, we develop an infrastructure that 1) automatically generates queries to probe LLMs and 2) evaluates answers to these queries using multiple LLM-as-a-judge setups and prompts. For 1), our prompt creation pipeline samples the space of patient demographics, histories, disorders, and writing styles to create realistic questions that we subsequently use to prompt LLMs. In 2), our evaluation pipeline provides hallucination and omission detection using LLM-as-a-judge as well as agentic workflows, in addition to LLM-as-a-judge treatment category detectors. As a baseline study, we perform two case studies on inter-LLM agreement and the impact of varying the answering and evaluation LLMs. We find that LLM annotators exhibit low agreement scores (average Cohen's Kappa $\kappa=0.118$), and only specific (answering, evaluation) LLM pairs yield statistically significant differences across writing styles, genders, and races. We recommend that studies using LLM evaluation use multiple LLMs as evaluators in order to avoid arriving at statistically significant but non-generalizable results, particularly in the absence of ground-truth data. We also suggest publishing inter-LLM agreement metrics for transparency. Our code and dataset are available here: https://github.com/BBN-E/medic-neurips-2025-demo.
KFUPM King Fahd Univeris
Why we think this paper is great for you:
This study directly investigates the alignment of responsible AI values between LLMs and human judgment, which is central to your focus on ethical AI development. It provides insights into how well AI systems reflect human ethical considerations.
This study directly investigates the alignment of responsible AI values between LLMs and human judgment, which is central to your focus on ethical AI development. It provides insights into how well AI systems reflect human ethical considerations.
Abstract
Large Language Models (LLMs) are increasingly employed in software engineering tasks such as requirements elicitation, design, and evaluation, raising critical questions regarding their alignment with human judgments on responsible AI values. This study investigates how closely LLMs' value preferences align with those of two human groups: a US-representative sample and AI practitioners. We evaluate 23 LLMs across four tasks: (T1) selecting key responsible AI values, (T2) rating their importance in specific contexts, (T3) resolving trade-offs between competing values, and (T4) prioritizing software requirements that embody those values. The results show that LLMs generally align more closely with AI practitioners than with the US-representative sample, emphasizing fairness, privacy, transparency, safety, and accountability. However, inconsistencies appear between the values that LLMs claim to uphold (Tasks 1-3) and the way they prioritize requirements (Task 4), revealing gaps in faithfulness between stated and applied behavior. These findings highlight the practical risk of relying on LLMs in requirements engineering without human oversight and motivate the need for systematic approaches to benchmark, interpret, and monitor value alignment in AI-assisted software development.
Large Language Models (LLMs) are increasingly employed in software engineering tasks such as requirements elicitation, design, and evaluation, raising critical questions regarding their alignment with human judgments on responsible AI values. This study investigates how closely LLMs' value preferences align with those of two human groups: a US-representative sample and AI practitioners. We evaluate 23 LLMs across four tasks: (T1) selecting key responsible AI values, (T2) rating their importance in specific contexts, (T3) resolving trade-offs between competing values, and (T4) prioritizing software requirements that embody those values. The results show that LLMs generally align more closely with AI practitioners than with the US-representative sample, emphasizing fairness, privacy, transparency, safety, and accountability. However, inconsistencies appear between the values that LLMs claim to uphold (Tasks 1-3) and the way they prioritize requirements (Task 4), revealing gaps in faithfulness between stated and applied behavior. These findings highlight the practical risk of relying on LLMs in requirements engineering without human oversight and motivate the need for systematic approaches to benchmark, interpret, and monitor value alignment in AI-assisted software development.
University of Virginia
Why we think this paper is great for you:
This paper directly addresses the crucial need for transparency in AI tools, specifically within human resources management, which aligns with your interest in AI and data transparency. It explores how to make black-boxed systems more understandable in a high-stakes context.
This paper directly addresses the crucial need for transparency in AI tools, specifically within human resources management, which aligns with your interest in AI and data transparency. It explores how to make black-boxed systems more understandable in a high-stakes context.
Abstract
AI tools are proliferating in human resources management (HRM) and recruiting, helping to mediate access to the labor market. As these systems spread, profession-specific transparency needs emerging from black-boxed systems in HRM move into focus. Prior work often frames transparency technically or abstractly, but we contend AI transparency is a social project shaped by materials, meanings, and competencies of practice. This paper introduces the Talent Acquisition and Recruiting AI (TARAI) Index, situating AI systems within the social practice of recruiting by examining product functionality, claims, assumptions, and AI clarity. Built through an iterative, mixed-methods process, the database demonstrates how transparency emerges: not as a fixed property, but as a dynamic outcome shaped by professional practices, interactions, and competencies. By centering social practice, our work offers a grounded, actionable approach to understanding and articulating AI transparency in HR and provides a blueprint for participatory database design for contextual transparency in professional practice.
AI tools are proliferating in human resources management (HRM) and recruiting, helping to mediate access to the labor market. As these systems spread, profession-specific transparency needs emerging from black-boxed systems in HRM move into focus. Prior work often frames transparency technically or abstractly, but we contend AI transparency is a social project shaped by materials, meanings, and competencies of practice. This paper introduces the Talent Acquisition and Recruiting AI (TARAI) Index, situating AI systems within the social practice of recruiting by examining product functionality, claims, assumptions, and AI clarity. Built through an iterative, mixed-methods process, the database demonstrates how transparency emerges: not as a fixed property, but as a dynamic outcome shaped by professional practices, interactions, and competencies. By centering social practice, our work offers a grounded, actionable approach to understanding and articulating AI transparency in HR and provides a blueprint for participatory database design for contextual transparency in professional practice.
Tsinghua University, Mon
Why we think this paper is great for you:
This paper explores the effects of generative AI agents on ethical deliberation, particularly in collaborative reasoning scenarios. It offers valuable insights into the moral implications of integrating AI into human decision-making processes.
This paper explores the effects of generative AI agents on ethical deliberation, particularly in collaborative reasoning scenarios. It offers valuable insights into the moral implications of integrating AI into human decision-making processes.
Abstract
Generative AI is increasingly positioned as a peer in collaborative learning, yet its effects on ethical deliberation remain unclear. We report a between-subjects experiment with university students (N=217) who discussed an autonomous-vehicle dilemma in triads under three conditions: human-only control, supportive AI teammate, or contrarian AI teammate. Using moral foundations lexicons, argumentative coding from the augmentative knowledge construction framework, semantic trajectory modelling with BERTopic and dynamic time warping, and epistemic network analysis, we traced how AI personas reshape moral discourse. Supportive AIs increased grounded/qualified claims relative to control, consolidating integrative reasoning around care/fairness, while contrarian AIs modestly broadened moral framing and sustained value pluralism. Both AI conditions reduced thematic drift compared with human-only groups, indicating more stable topical focus. Post-discussion justification complexity was only weakly predicted by moral framing and reasoning quality, and shifts in final moral decisions were driven primarily by participants' initial stance rather than condition. Overall, AI teammates altered the process, the distribution and connection of moral frames and argument quality, more than the outcome of moral choice, highlighting the potential of generative AI agents as teammates for eliciting reflective, pluralistic moral reasoning in collaborative learning.
Generative AI is increasingly positioned as a peer in collaborative learning, yet its effects on ethical deliberation remain unclear. We report a between-subjects experiment with university students (N=217) who discussed an autonomous-vehicle dilemma in triads under three conditions: human-only control, supportive AI teammate, or contrarian AI teammate. Using moral foundations lexicons, argumentative coding from the augmentative knowledge construction framework, semantic trajectory modelling with BERTopic and dynamic time warping, and epistemic network analysis, we traced how AI personas reshape moral discourse. Supportive AIs increased grounded/qualified claims relative to control, consolidating integrative reasoning around care/fairness, while contrarian AIs modestly broadened moral framing and sustained value pluralism. Both AI conditions reduced thematic drift compared with human-only groups, indicating more stable topical focus. Post-discussion justification complexity was only weakly predicted by moral framing and reasoning quality, and shifts in final moral decisions were driven primarily by participants' initial stance rather than condition. Overall, AI teammates altered the process, the distribution and connection of moral frames and argument quality, more than the outcome of moral choice, highlighting the potential of generative AI agents as teammates for eliciting reflective, pluralistic moral reasoning in collaborative learning.
University of Texas at D
Why we think this paper is great for you:
This research explores how revealing AI reasoning impacts human trust and knowledge, which is directly relevant to your interest in AI transparency. It provides a nuanced perspective on the benefits and potential drawbacks of making AI processes more understandable.
This research explores how revealing AI reasoning impacts human trust and knowledge, which is directly relevant to your interest in AI transparency. It provides a nuanced perspective on the benefits and potential drawbacks of making AI processes more understandable.
Abstract
Effective human-AI collaboration requires humans to accurately gauge AI capabilities and calibrate their trust accordingly. Humans often have context-dependent private information, referred to as Unique Human Knowledge (UHK), that is crucial for deciding whether to accept or override AI's recommendations. We examine how displaying AI reasoning affects trust and UHK utilization through a pre-registered, incentive-compatible experiment (N = 752). We find that revealing AI reasoning, whether brief or extensive, acts as a powerful persuasive heuristic that significantly increases trust and agreement with AI recommendations. Rather than helping participants appropriately calibrate their trust, this transparency induces over-trust that crowds out UHK utilization. Our results highlight the need for careful consideration when revealing AI reasoning and call for better information design in human-AI collaboration systems.
Effective human-AI collaboration requires humans to accurately gauge AI capabilities and calibrate their trust accordingly. Humans often have context-dependent private information, referred to as Unique Human Knowledge (UHK), that is crucial for deciding whether to accept or override AI's recommendations. We examine how displaying AI reasoning affects trust and UHK utilization through a pre-registered, incentive-compatible experiment (N = 752). We find that revealing AI reasoning, whether brief or extensive, acts as a powerful persuasive heuristic that significantly increases trust and agreement with AI recommendations. Rather than helping participants appropriately calibrate their trust, this transparency induces over-trust that crowds out UHK utilization. Our results highlight the need for careful consideration when revealing AI reasoning and call for better information design in human-AI collaboration systems.
Data Bias
Meta
Abstract
As we exhaust methods that reduces variance without introducing bias, reducing variance in experiments often requires accepting some bias, using methods like winsorization or surrogate metrics. While this bias-variance tradeoff can be optimized for individual experiments, bias may accumulate over time, raising concerns for long-term optimization. We analyze whether bias is ever acceptable when it can accumulate, and show that a bias-variance tradeoff persists in long-term settings. Improving signal-to-noise remains beneficial, even if it introduces bias. This implies we should shift from thinking there is a single ``correct'', unbiased metric to thinking about how to make the best estimates and decisions when better precision can be achieved at the expense of bias. Furthermore, our model adds nuance to previous findings that suggest less stringent launch criterion leads to improved gains. We show while this is beneficial when the system is far from the optimum, more stringent launch criterion is preferable as the system matures.
As we exhaust methods that reduces variance without introducing bias, reducing variance in experiments often requires accepting some bias, using methods like winsorization or surrogate metrics. While this bias-variance tradeoff can be optimized for individual experiments, bias may accumulate over time, raising concerns for long-term optimization. We analyze whether bias is ever acceptable when it can accumulate, and show that a bias-variance tradeoff persists in long-term settings. Improving signal-to-noise remains beneficial, even if it introduces bias. This implies we should shift from thinking there is a single ``correct'', unbiased metric to thinking about how to make the best estimates and decisions when better precision can be achieved at the expense of bias. Furthermore, our model adds nuance to previous findings that suggest less stringent launch criterion leads to improved gains. We show while this is beneficial when the system is far from the optimum, more stringent launch criterion is preferable as the system matures.
Data Fairness
Concordia University, FAI
Abstract
This paper introduces a theoretical framework to resolve a central paradox in modern machine learning: When is it better to use less data? This question has become critical as classical scaling laws suggesting ``more is more'' (Sun et al., 2025) are challenged by methods like LIMO (``less is more'') and s1 (Ye et al., 2025; Muenighoff et al., 2025), which achieve superior performance with small, aggressively curated datasets. Here, we study data curation strategies where an imperfect oracle selects the training examples according to their difficulty and correctness. Our results provide exact scaling law curves for test error under both label-agnostic and label-aware curation rules, revealing when and why keeping only a subset of data can improve generalization. In contrast to classical scaling laws, we show that under certain conditions, small curated datasets can outperform full datasets, and we provide analytical conditions for this by deriving precise phase transition curves tied to data size and quality. We validate these theoretical claims with empirical results on ImageNet, confirming our predictions about when curation improves accuracy and can even mitigate model collapse. Furthermore, our framework provides a principled explanation for the contradictory curation strategies recently observed in LLM mathematical reasoning.
This paper introduces a theoretical framework to resolve a central paradox in modern machine learning: When is it better to use less data? This question has become critical as classical scaling laws suggesting ``more is more'' (Sun et al., 2025) are challenged by methods like LIMO (``less is more'') and s1 (Ye et al., 2025; Muenighoff et al., 2025), which achieve superior performance with small, aggressively curated datasets. Here, we study data curation strategies where an imperfect oracle selects the training examples according to their difficulty and correctness. Our results provide exact scaling law curves for test error under both label-agnostic and label-aware curation rules, revealing when and why keeping only a subset of data can improve generalization. In contrast to classical scaling laws, we show that under certain conditions, small curated datasets can outperform full datasets, and we provide analytical conditions for this by deriving precise phase transition curves tied to data size and quality. We validate these theoretical claims with empirical results on ImageNet, confirming our predictions about when curation improves accuracy and can even mitigate model collapse. Furthermore, our framework provides a principled explanation for the contradictory curation strategies recently observed in LLM mathematical reasoning.
Data Representation
Stanford University and
Abstract
This paper outlines a grammar of data analysis, as distinct from grammars of data manipulation, in which the primitives are metrics and dimensions. We describe a Python implementation of this grammar called Meterstick, which is agnostic to the underlying data source, which may be a DataFrame or a SQL database.
This paper outlines a grammar of data analysis, as distinct from grammars of data manipulation, in which the primitives are metrics and dimensions. We describe a Python implementation of this grammar called Meterstick, which is agnostic to the underlying data source, which may be a DataFrame or a SQL database.
University of Illinois at
Abstract
Unstructured data, in the form of text, images, video, and audio, is produced at exponentially higher rates. In tandem, machine learning (ML) methods have become increasingly powerful at analyzing unstructured data. Modern ML methods can now detect objects in images, understand actions in videos, and even classify complex legal texts based on legal intent. Combined, these trends make it increasingly feasible for analysts and researchers to automatically understand the "real world." However, there are major challenges in deploying these techniques: 1) executing queries efficiently given the expense of ML methods, 2) expressing queries over bespoke forms of data, and 3) handling errors in ML methods. In this monograph, we discuss challenges and advances in data management systems for unstructured data using ML, with a particular focus on video analytics. Using ML to answer queries introduces new challenges.First, even turning user intent into queries can be challenging: it is not obvious how to express a query of the form "select instances of cars turning left." Second, ML models can be orders of magnitude more expensive compared to processing traditional structured data. Third, ML models and the methods to accelerate analytics with ML models can be error-prone. Recent work in the data management community has aimed to address all of these challenges. Users can now express queries via user-defined functions, opaquely through standard structured schemas, and even by providing examples. Given a query, recent work focuses on optimizing queries by approximating expensive "gold" methods with varying levels of guarantees. Finally, to handle errors in ML models, recent work has focused on applying outlier and drift detection to data analytics with ML.
Unstructured data, in the form of text, images, video, and audio, is produced at exponentially higher rates. In tandem, machine learning (ML) methods have become increasingly powerful at analyzing unstructured data. Modern ML methods can now detect objects in images, understand actions in videos, and even classify complex legal texts based on legal intent. Combined, these trends make it increasingly feasible for analysts and researchers to automatically understand the "real world." However, there are major challenges in deploying these techniques: 1) executing queries efficiently given the expense of ML methods, 2) expressing queries over bespoke forms of data, and 3) handling errors in ML methods. In this monograph, we discuss challenges and advances in data management systems for unstructured data using ML, with a particular focus on video analytics. Using ML to answer queries introduces new challenges.First, even turning user intent into queries can be challenging: it is not obvious how to express a query of the form "select instances of cars turning left." Second, ML models can be orders of magnitude more expensive compared to processing traditional structured data. Third, ML models and the methods to accelerate analytics with ML models can be error-prone. Recent work in the data management community has aimed to address all of these challenges. Users can now express queries via user-defined functions, opaquely through standard structured schemas, and even by providing examples. Given a query, recent work focuses on optimizing queries by approximating expensive "gold" methods with varying levels of guarantees. Finally, to handle errors in ML models, recent work has focused on applying outlier and drift detection to data analytics with ML.
AI Bias
Syracuse University, USA
Abstract
Popular discourses are thick with narratives of generative AI's problematic functions and outcomes, yet there is little understanding of how non-experts consider AI activities to constitute bad behavior. This study starts to bridge that gap through inductive analysis of interviews with non-experts (N = 28) focusing on large-language models in general and their bad behavior, specifically. Results suggest bad behaviors are not especially salient when people discuss AI generally but the notion of AI behaving badly is easily engaged when prompted, and bad behavior becomes even more salient when evaluating specific AI behaviors. Types of observed behaviors considered bad mostly align with their inspiring moral foundations; across all observed behaviors, some variations on non-performance and social discordance were present. By scaffolding findings at the intersections of moral foundations theory, construal level theory, and moral dyadism, a tentative framework for considering AI bad behavior is proposed.
Popular discourses are thick with narratives of generative AI's problematic functions and outcomes, yet there is little understanding of how non-experts consider AI activities to constitute bad behavior. This study starts to bridge that gap through inductive analysis of interviews with non-experts (N = 28) focusing on large-language models in general and their bad behavior, specifically. Results suggest bad behaviors are not especially salient when people discuss AI generally but the notion of AI behaving badly is easily engaged when prompted, and bad behavior becomes even more salient when evaluating specific AI behaviors. Types of observed behaviors considered bad mostly align with their inspiring moral foundations; across all observed behaviors, some variations on non-performance and social discordance were present. By scaffolding findings at the intersections of moral foundations theory, construal level theory, and moral dyadism, a tentative framework for considering AI bad behavior is proposed.
University of Illinois
Abstract
In many real-life settings, algorithms play the role of assistants, while humans ultimately make the final decision. Often, algorithms specifically act as curators, narrowing down a wide range of options into a smaller subset that the human picks between: consider content recommendation or chatbot responses to questions with multiple valid answers. Crucially, humans may not know their own preferences perfectly either, but instead may only have access to a noisy sampling over preferences. Algorithms can assist humans by curating a smaller subset of items, but must also face the challenge of misalignment: humans may have different preferences from each other (and from the algorithm), and the algorithm may not know the exact preferences of the human they are facing at any point in time. In this paper, we model and theoretically study such a setting. Specifically, we show instances where humans benefit by collaborating with a misaligned algorithm. Surprisingly, we show that humans gain more utility from a misaligned algorithm (which makes different mistakes) than from an aligned algorithm. Next, we build on this result by studying what properties of algorithms maximize human welfare when the goals could be either utilitarian welfare or ensuring all humans benefit. We conclude by discussing implications for designers of algorithmic tools and policymakers.
In many real-life settings, algorithms play the role of assistants, while humans ultimately make the final decision. Often, algorithms specifically act as curators, narrowing down a wide range of options into a smaller subset that the human picks between: consider content recommendation or chatbot responses to questions with multiple valid answers. Crucially, humans may not know their own preferences perfectly either, but instead may only have access to a noisy sampling over preferences. Algorithms can assist humans by curating a smaller subset of items, but must also face the challenge of misalignment: humans may have different preferences from each other (and from the algorithm), and the algorithm may not know the exact preferences of the human they are facing at any point in time. In this paper, we model and theoretically study such a setting. Specifically, we show instances where humans benefit by collaborating with a misaligned algorithm. Surprisingly, we show that humans gain more utility from a misaligned algorithm (which makes different mistakes) than from an aligned algorithm. Next, we build on this result by studying what properties of algorithms maximize human welfare when the goals could be either utilitarian welfare or ensuring all humans benefit. We conclude by discussing implications for designers of algorithmic tools and policymakers.
Data Ethics
University of Surrey
Abstract
Crowdsourced data supports real-time decision-making but faces challenges like misinformation, errors, and contributor power concentration. This study systematically examines trust management practices across platforms categorised as Volunteered Geographic Information, Wiki Ecosystems, Social Media, Mobile Crowdsensing, and Specialised Review and Environmental Crowdsourcing. Identified strengths include automated moderation and community validation, while limitations involve rapid data influx, niche oversight gaps, opaque trust metrics, and elite dominance. Proposed solutions incorporate advanced AI tools, transparent reputation metrics, decentralised moderation, structured community engagement, and a ``soft power'' strategy, aiming to equitably distribute decision-making authority and enhance overall data reliability.
Crowdsourced data supports real-time decision-making but faces challenges like misinformation, errors, and contributor power concentration. This study systematically examines trust management practices across platforms categorised as Volunteered Geographic Information, Wiki Ecosystems, Social Media, Mobile Crowdsensing, and Specialised Review and Environmental Crowdsourcing. Identified strengths include automated moderation and community validation, while limitations involve rapid data influx, niche oversight gaps, opaque trust metrics, and elite dominance. Proposed solutions incorporate advanced AI tools, transparent reputation metrics, decentralised moderation, structured community engagement, and a ``soft power'' strategy, aiming to equitably distribute decision-making authority and enhance overall data reliability.
Interests not found
We did not find any papers that match the below interests.
Try other terms also consider if the content exists in arxiv.org.
- Data Transparency
You can edit or add more interests any time.
Help us improve your experience!
This project is on its early stages your feedback can be pivotal on the future of the project.
Let us know what you think about this week's papers and suggestions!
Give Feedback