Proposes a new method to measure and avoid harm in machine learning decisions.
problem Measuring and avoiding harm in machine learning algorithms.
method Formal definition of harm and benefit using causal models, counterfactual objective functions.
result Demonstrates that standard machine learning methods can lead to harmful policies under distributional shifts.
Prediction models can harm patients even when accurate, leading to self-fulfilling prophecies.
problem Prediction models can lead to harmful decisions that worsen patient outcomes.
method Formal characterization of harmful prediction models and analysis of their impact.
result Well-calibrated models are ineffective for decision-making as they do not change the data distribution.
Improving cancer treatment decisions requires considering causal effects, not just model accuracy.
problem Cancer outcome prediction models may cause harm when used for treatment decisions.
method Explains the importance of considering causal effects in model validation and provides guidelines.
result Building and validating models that are useful for decision making requires considering causal effects.
New framework shows algorithmic recourse can be harmful.
problem Risks of providing algorithmic recourse in binary classification.
method Abstract learning-theoretic framework comparing risks with and without recourse.
result Providing recourse can be harmful, increasing class uncertainty and mistakes.
Neural Networks are being integrated into safety critical systems, e.g., perception systems for autonomous vehicles, which require trained networks to perform safely in novel scenarios. It is challenging to verify neural networks because their decisions are not explainable, they cannot be exhaustively tested, and finit…
Consequential decision-making typically incentivizes individuals to behave strategically, tailoring their behavior to the specifics of the decision rule. A long line of work has therefore sought to counteract strategic behavior by designing more conservative decision boundaries in an effort to increase robustness to th…
New mechanism designs regulate herding in financial markets.
problem Herding causes irrational market decisions and volatility.
method A trilateral game framework based on optimal control theory.
result Effective mechanisms improve social welfare.
Contrastive learning harms minority group representations, affecting downstream tasks.
problem Representation harm in contrastive learning, especially affecting minority groups.
method Causal mediation analysis and stochastic block model explanation.
result Representation harm in contrastive learning is partly responsible for allocation harm in downstream tasks.
Cold posteriors in BNNs harm performance, likely due to incorrect likelihood.
problem Cold posteriors in Bayesian neural networks degrade performance.
method Developed a generative model explaining cold posteriors and matched it to the tempered likelihoods.
result Cold posteriors are a result of using the wrong likelihood for image classification datasets.
While harms of allocation have been increasingly studied as part of the subfield of algorithmic fairness, harms of representation have received considerably less attention. In this paper, we formalize two notions of stereotyping and show how they manifest in later allocative harms within the machine learning pipeline. …
AI systems that explain their decisions can be monitored for harmful intentions.
problem Monitoring AI systems' decision-making processes for harmful intentions is imperfect and can miss some misbehavior.
method Monitoring the chain of thought (CoT) of AI systems that communicate in human language.
result CoT monitoring is a promising but fragile approach to AI safety.
COCA refactors training data to identify and erase unsafe concepts in LLMs.
problem Identifying and erasing unsafe concepts in Large Language Models (LLMs) for safety alignment.
method Concept Concentration (COCA) refactors training data with an explicit reasoning process to identify and erase unsafe concepts.
result COCA significantly reduces both in-distribution and out-of-distribution jailbreak success rates while maintaining strong performance on regular tasks.
Aims to create safe reinforcement learning policies by considering individual harm.
problem Optimal policies for a population may harm certain individuals.
method Formalizes individual harm, proposes a two-stage procedure, and establishes finite-sample properties.
result Learned policies maximize expected return while minimizing harm.
Researchers classify harmful structures on unimodular Lie groups.
problem Characterizing harmful structures on Lie groups.
method Analyzing left-invariant structures and using Clifford multiplication.
result Classification of harmful structures on unimodular Lie groups of dimension ≤ 4.
In this paper we propose an algorithm that builds sparse decision DAGs (directed acyclic graphs) from a list of base classifiers provided by an external learning method such as AdaBoost. The basic idea is to cast the DAG design task as a Markov decision process. Each instance can decide to use or to skip each base clas…
New method stops experiments early for harm in diverse groups.
problem Early stopping of experiments for harmful treatment effects in diverse populations.
method Causal machine learning approach (CLASH) for early stopping.
result CLASH effectively stops experiments early for harmful treatment effects in diverse groups.
Machine learning algorithms are now frequently used in sensitive contexts that substantially affect the course of human lives, such as credit lending or criminal justice. This is driven by the idea that `objective' machines base their decisions solely on facts and remain unaffected by human cognitive biases, discrimina…
Develops a support-aware framework for reserve-policy selection in advertising markets.
problem Log-based reserve-price evaluation risks weak support and subgroup harm.
method Support-aware offline decision framework converting logged evidence into certified policies.
result Preserves the best gate-passing policy while eliminating only policies with certified regret.
Black box machine learning models are currently being used for high stakes decision-making throughout society, causing problems throughout healthcare, criminal justice, and in other domains. People have hoped that creating methods for explaining these black box models will alleviate some of these problems, but trying t…
Theorem. Let M be a compact, connected, oriented smooth Riemannian n-manifold with non-empty boundary. Then the cohomology of the complex (Harm*(M),d) of harmonic forms on M is given by the direct sum H^p(Harm*(M),d) = H^p(M;R) + H^(p-1)(M;R) for p=0,1,...,n. When M is a closed manifold, a form is harmonic if and only …
Proposes a framework to create fair IDRs by enforcing demographic parity constraints.
problem Discrimination in IDRs trained on biased data.
method Incorporates DP and CDP constraints into IDR estimation.
result Theoretically optimal IDRs can be efficiently obtained through perturbations.
New method estimates harmful instances in GANs for better model performance.
problem Estimating harmful instances in GANs for improved model performance.
method Uses Jacobian of generator's loss to trace harmfulness of instances.
result Correctly inferred changes in GAN evaluation metrics and improved model performance.
Study shows harmful overfitting in Sobolev spaces even as training data grows.
problem Harmful overfitting in Sobolev spaces under noisy conditions.
method Geometric argument using Sobolev inequalities.
result Approximately norm-minimizing interpolators exhibit harmful overfitting.
The paper tackles optimal policy learning with asymmetric counterfactual utilities in healthcare decisions.
problem Learning optimal policies from observed data with asymmetric counterfactual utilities.
method The approach involves identifying and minimizing the maximum expected utility loss using statistical decision theory and solving intermediate classification problems.
result One can learn minimax loss decision rules from observed data.
Fuses ITRs for primary and secondary outcomes to minimize harm.
problem Learn an ITR maximizing primary outcome while minimizing harm to secondary outcomes.
method Introduces fusion penalty to encourage similar recommendations for different outcomes. Two algorithms estimate the ITR using surrogate loss functions.
result Agreement rate between primary and secondary optimal ITRs converges faster than ignoring secondary outcomes.
As machine learning (ML) increasingly affects people and society, awareness of its potential unwanted consequences has also grown. To anticipate, prevent, and mitigate undesirable downstream consequences, it is critical that we understand when and how harm might be introduced throughout the ML life cycle. In this paper…
Common fairness definitions in machine learning focus on balancing notions of disparity and utility. In this work, we study fairness in the context of risk disparity among sub-populations. We are interested in learning models that minimize performance discrepancies across sensitive groups without causing unnecessary ha…
New method handles uncertainty in causal effect estimation for better decision-making.
problem Handling uncertainty in causal effect estimation, especially in high-dimensional data and covariate shift.
method Integrates uncertainty estimation into neural network methods for individual-level causal estimates.
result Uncertainty-aware methods improve decision-making by alerting when predictions are not reliable.
Optimal allocation of human effort to correct AI assessments in decision-making.
problem How to allocate costly human effort to correct noisy or biased AI-generated assessments.
method Decision-theoretic framework treating AI assessments as signals and human judgments as costly information. Developed estimation procedures under nonparametric and linear models.
result Our approach substantially outperforms LLM-only predictions and achieves performance comparable to full human review while using only 20-30% of the human information.
Improves GAN performance by identifying and removing harmful training instances.
problem Challenges in applying previous instance removal methods to GANs.
method Proposes influence estimation methods using Jacobian of generator's gradient and instance evaluation scheme based on GAN evaluation metrics.
result Removing harmful instances significantly improves GAN performance.
Bayesian approach quantifies uncertainty in LLM evaluations.
problem Statistical uncertainty in evaluating LLM behavior.
method Bayesian evaluation of LLM behavior using probabilistic text generation strategies.
result Bayesian approach provides useful uncertainty quantification about LLM behavior.
New framework identifies worst-case shifts for predictive resource allocation models.
problem Identifying harmful shifts in predictive models for resource allocation.
method Hierarchical model structure and submodular optimization for worst-case loss.
result Empirical evidence shows divergent worst-case shifts identified by different metrics.
Bayesian system helps identify and thwart terrorist plans.
problem Identifying and thwarting terrorist plans before they occur.
method Develops a Bayesian decision support system to integrate member and group activities.
result Estimates the combined threat posed by a terrorist group.
Today, artificial intelligence systems driven by machine learning algorithms can be in a position to take important, and sometimes legally binding, decisions about our everyday lives. In many cases, however, these systems and their actions are neither regulated nor certified. To help counter the potential harm that suc…
DFFL tackles federated learning with heterogeneous objectives and constraints.
problem Federated learning with clients having different objectives and feasible regions.
method Derived heterogeneity bounds for cost-vector distances and support-function/shape-distance terms. Lifted pointwise bounds to local-versus-federated excess-risk comparison.
result Federation is beneficial when the statistical advantage of pooling exceeds a client-specific heterogeneity penalty.
The study finds that memorization is necessary or harmful depending on the prior distribution and noise level.
problem The impact of memorization on generalization in overparameterized models.
method An overparameterized linear model with general priors in a Bayesian setup.
result Explicit conditions for optimal generalization based on the prior distribution and noise level.
In many real-world reinforcement learning (RL) problems, besides optimizing the main objective function, an agent must concurrently avoid violating a number of constraints. In particular, besides optimizing performance it is crucial to guarantee the safety of an agent during training as well as deployment (e.g. a robot…
Bayesian optimization enhanced with conformal prediction for better outcome reliability.
problem Uncertainty and model misspecification in Bayesian optimization.
method Conformal prediction to provide coverage guarantees and Bayesian optimization to select queries.
result Significant improvement in query coverage without sacrificing sample-efficiency.
Better use of unlabelled data improves Bayesian active learning models.
problem Neglecting unlabelled data harms predictive performance and data acquisition decisions.
method A simple framework for semi-supervised Bayesian active learning.
result The proposed framework produces better models than conventional approaches.
We study the problem of learning personalized decision policies from observational data while accounting for possible unobserved confounding. Previous approaches, which assume unconfoundedness, i.e., that no unobserved confounders affect both the treatment assignment as well as outcome, can lead to policies that introd…
This work advances collaborative decision making by combining human and AI strengths in uncertainty quantification.
problem Current AI lacks robust decision-making capabilities under uncertainty, especially in high-stakes contexts.
method Introduces Human AI Collaborative Uncertainty Quantification (HACUQ) framework, formalizing AI-human collaboration and developing calibration algorithms.
result Optimal collaborative prediction sets follow a two-threshold structure, and online adaptation algorithms can adapt to evolving human behavior.
Diabetes mellitus is a common disease of human body caused by a group of metabolic disorders where the sugar levels over a prolonged period is very high. It affects different organs of the human body which thus harm a large number of the body's system, in particular the blood veins and nerves. Early prediction in such …
New framework assesses LLM security risks in BFSI.
problem Lack of domain-specific security evaluation for LLMs in BFSI.
method Risk-aware evaluation framework combining taxonomy, automated red-teaming, and ensemble judging.
result Higher decoding stochasticity and adaptive interaction lead to more severe disclosures.
A primary goal of computational phenotype research is to conduct medical diagnosis. In hospital, physicians rely on massive clinical data to make diagnosis decisions, among which laboratory tests are one of the most important resources. However, the longitudinal and incomplete nature of laboratory test data casts a sig…
The study evaluates AI model performance measures for medical use.
problem Selecting appropriate performance measures for AI models in medical practice.
method Assessed 32 performance measures across five domains for binary outcomes.
result 17 measures are both proper and reflect decision-analytic performance.
This paper introduces a Bayesian framework for optimizing online experiments to maximize profit.
problem Statistical flaws and reliance on proxy metrics in A/B tests compromise their effectiveness.
method Hierarchical Bayesian model for estimating conversion probability and monetary value, decision-theoretic stopping rule.
result The framework ensures experiments conclude when no variant offers a significant profit improvement, conserving resources.
Machine learning algorithms increasingly influence our decisions and interact with us in all parts of our daily lives. Therefore, just as we consider the safety of power plants, highways, and a variety of other engineered socio-technical systems, we must also take into account the safety of systems involving machine le…
New attacks reduce bad queries in black-box classifiers, improving effectiveness.
problem Evasion attacks often issue many bad queries, which are costly.
method Design new attacks that minimize bad queries while increasing non-bad queries.
result Reduces the number of bad queries by 1.5-7.3 times, improving attack effectiveness.