Christoph Kern

Jan Simson, Florian Pfisterer, Christoph Kern

A vast number of systems across the world use algorithmic decision making (ADM) to (partially) automate decisions that have previously been made by humans. The downstream effects of ADM systems critically depend on the decisions made during a systems' design, implementation, and evaluation, as biases in data can be mitigated or reinforced along the modeling pipeline. Many of these decisions are made implicitly, without knowing exactly how they will influence the final system. To study this issue, we draw on insights from the field of psychology and introduce the method of multiverse analysis for algorithmic fairness. In our proposed method, we turn implicit decisions during design and evaluation into explicit ones and demonstrate their fairness implications. By combining decisions, we create a grid of all possible "universes" of decision combinations. For each of these universes, we compute metrics of fairness and performance. Using the resulting dataset, one can investigate the variability and robustness of fairness scores and see how and which decisions impact fairness. We demonstrate how multiverse analyses can be used to better understand fairness implications of design and evaluation decisions using an exemplary case study of predicting public health care coverage for vulnerable populations. Our results highlight how decisions regarding the evaluation of a system can lead to vastly different fairness metrics for the same model. This is problematic, as a nefarious actor could optimise or "hack" a fairness metric to portray a discriminating model as fair merely by changing how it is evaluated. We illustrate how a multiverse analysis can help to address this issue.

Christoph Kern, Stephanie Eckman, Jacob Beck et al.

When training data are collected from human annotators, the design of the annotation instrument, the instructions given to annotators, the characteristics of the annotators, and their interactions can impact training data. This study demonstrates that design choices made when creating an annotation instrument also impact the models trained on the resulting annotations. We introduce the term annotation sensitivity to refer to the impact of annotation data collection methods on the annotations themselves and on downstream model performance and predictions. We collect annotations of hate speech and offensive language in five experimental conditions of an annotation instrument, randomly assigning annotators to conditions. We then fine-tune BERT models on each of the five resulting datasets and evaluate model performance on a holdout portion of each condition. We find considerable differences between the conditions for 1) the share of hate speech/offensive language annotations, 2) model performance, 3) model predictions, and 4) model learning curves. Our results emphasize the crucial role played by the annotation instrument which has received little attention in the machine learning literature. We call for additional research into how and why the instrument impacts the annotations to inform the development of best practices in instrument design.

Edoardo Debenedetti, Ilia Shumailov, Tianqi Fan et al. · deepmind, eth-zurich

Large Language Models (LLMs) are increasingly deployed in agentic systems that interact with an untrusted environment. However, LLM agents are vulnerable to prompt injection attacks when handling untrusted data. In this paper we propose CaMeL, a robust defense that creates a protective system layer around the LLM, securing it even when underlying models are susceptible to attacks. To operate, CaMeL explicitly extracts the control and data flows from the (trusted) query; therefore, the untrusted data retrieved by the LLM can never impact the program flow. To further improve security, CaMeL uses a notion of a capability to prevent the exfiltration of private data over unauthorized data flows by enforcing security policies when tools are called. We demonstrate effectiveness of CaMeL by solving $77\%$ of tasks with provable security (compared to $84\%$ with an undefended system) in AgentDojo. We release CaMeL at https://github.com/google-research/camel-prompt-injection.

Patrick Kaiser, Christoph Kern, David Rügamer

Both industry and academia have made considerable progress in developing trustworthy and responsible machine learning (ML) systems. While critical concepts like fairness and explainability are often addressed, the safety of systems is typically not sufficiently taken into account. By viewing data-driven decision systems as socio-technical systems, we draw on the uncertainty in ML literature to show how fairML systems can also be safeML systems. We posit that a fair model needs to be an uncertainty-aware model, e.g. by drawing on distributional regression. For fair decisions, we argue that a safe fail option should be used for individuals with uncertain categorization. We introduce semi-structured deep distributional regression as a modeling framework which addresses multiple concerns brought against standard ML models and show its use in a real-world example of algorithmic profiling of job seekers.

Robert Chew, Stephanie Eckman, Christoph Kern et al.

Supervised machine learning assumes that labeled data provide accurate measurements of the concepts models are meant to learn. Yet in practice, human labeling introduces systematic variation arising from ambiguous items, divergent interpretations, and simple mistakes. Machine learning research commonly treats all disagreement as noise, which obscures these distinctions and limits our understanding of what models actually learn. This paper reframes annotation as a measurement process and introduces a statistical framework for decomposing labeling outcomes into interpretable sources of variation: instance difficulty, annotator bias, situational noise, and relational alignment. The framework extends classical measurement-error models to accommodate both shared and individualized notions of truth, reflecting traditional and human label variation interpretations of error, and provides a diagnostic for assessing which regime better characterizes a given task. Applying the proposed model to a multi-annotator natural language inference dataset, we find empirical evidence for all four theorized components and demonstrate the effectiveness of our approach. We conclude with implications for data-centric machine learning and outline how this approach can guide the development of a more systematic science of labeling.

Unai Fischer-Abaigar, Emily Aiken, Christoph Kern et al.

Institutions increasingly use prediction to allocate scarce resources. From a design perspective, better predictions compete with other investments, such as expanding capacity or improving treatment quality. Here, the big question is not how to solve a specific allocation problem, but rather which problem to solve. In this work, we develop an empirical toolkit to help planners form principled answers to this question and quantify the bottom-line welfare impact of investments in prediction versus other policy levers such as expanding capacity and improving treatment quality. Applying our framework in two real-world case studies on German employment services and poverty targeting in Ethiopia, we illustrate how decision-makers can reliably derive context-specific conclusions about the relative value of prediction in their allocation problem. We make our software toolkit, rvp, and parts of our data available in order to enable future empirical work in this area.

Jan Simson, Alessandro Fabris, Christoph Kern

Data practices shape research and practice on fairness in machine learning (fair ML). Critical data studies offer important reflections and critiques for the responsible advancement of the field by highlighting shortcomings and proposing recommendations for improvement. In this work, we present a comprehensive analysis of fair ML datasets, demonstrating how unreflective yet common practices hinder the reach and reliability of algorithmic fairness findings. We systematically study protected information encoded in tabular datasets and their usage in 280 experiments across 142 publications. Our analyses identify three main areas of concern: (1) a \textbf{lack of representation for certain protected attributes} in both data and evaluations; (2) the widespread \textbf{exclusion of minorities} during data preprocessing; and (3) \textbf{opaque data processing} threatening the generalization of fairness research. By conducting exemplary analyses on the utilization of prominent datasets, we demonstrate how unreflective data decisions disproportionately affect minority groups, fairness metrics, and resultant model comparisons. Additionally, we identify supplementary factors such as limitations in publicly available data, privacy considerations, and a general lack of awareness, which exacerbate these challenges. To address these issues, we propose a set of recommendations for data usage in fairness research centered on transparency and responsible inclusion. This study underscores the need for a critical reevaluation of data practices in fair ML and offers directions to improve both the sourcing and usage of datasets.

Patrick Oliver Schenk, Christoph Kern

National Statistical Organizations (NSOs) increasingly draw on Machine Learning (ML) to improve the timeliness and cost-effectiveness of their products. When introducing ML solutions, NSOs must ensure that high standards with respect to robustness, reproducibility, and accuracy are upheld as codified, e.g., in the Quality Framework for Statistical Algorithms (QF4SA; Yung et al. 2022). At the same time, a growing body of research focuses on fairness as a pre-condition of a safe deployment of ML to prevent disparate social impacts in practice. However, fairness has not yet been explicitly discussed as a quality aspect in the context of the application of ML at NSOs. We employ Yung et al. (2022)'s QF4SA quality framework and present a mapping of its quality dimensions to algorithmic fairness. We thereby extend the QF4SA framework in several ways: we argue for fairness as its own quality dimension, we investigate the interaction of fairness with other dimensions, and we explicitly address data, both on its own and its interaction with applied methodology. In parallel with empirical illustrations, we show how our mapping can contribute to methodology in the domains of official statistics, algorithmic fairness, and trustworthy machine learning.

Jan Simson, Alessandro Fabris, Cosima Fröhner et al.

As machine learning (ML) systems are increasingly adopted in high-stakes decision-making domains, ensuring fairness in their outputs has become a central challenge. At the core of fair ML research are the datasets used to investigate bias and develop mitigation strategies. Yet, much of the existing work relies on a narrow selection of datasets--often arbitrarily chosen, inconsistently processed, and lacking in diversity--undermining the generalizability and reproducibility of results. To address these limitations, we present FairGround: a unified framework, data corpus, and Python package aimed at advancing reproducible research and critical data studies in fair ML classification. FairGround currently comprises 44 tabular datasets, each annotated with rich fairness-relevant metadata. Our accompanying Python package standardizes dataset loading, preprocessing, transformation, and splitting, streamlining experimental workflows. By providing a diverse and well-documented dataset corpus along with robust tooling, FairGround enables the development of fairer, more reliable, and more reproducible ML models. All resources are publicly available to support open and collaborative research.

Unai Fischer-Abaigar, Christoph Kern, Juan Carlos Perdomo

Machine learning is increasingly used in government programs to identify and support the most vulnerable individuals, prioritizing assistance for those at greatest risk over optimizing aggregate outcomes. This paper examines the welfare impacts of prediction in equity-driven contexts, and how they compare to other policy levers, such as expanding bureaucratic capacity. Through mathematical models and a real-world case study on long-term unemployment amongst German residents, we develop a comprehensive understanding of the relative effectiveness of prediction in surfacing the worst-off. Our findings provide clear analytical frameworks and practical, data-driven tools that empower policymakers to make principled decisions when designing these systems.

Stephanie Eckman, Bolei Ma, Christoph Kern et al.

Models trained on crowdsourced annotations may not reflect population views, if those who work as annotators do not represent the broader population. In this paper, we propose PAIR: Population-Aligned Instance Replication, a post-processing method that adjusts training data to better reflect target population characteristics without collecting additional annotations. Using simulation studies on offensive language and hate speech detection with varying annotator compositions, we show that non-representative pools degrade model calibration while leaving accuracy largely unchanged. PAIR corrects these calibration problems by replicating annotations from underrepresented annotator groups to match population proportions. We conclude with recommendations for improving the representativity of training data and model performance.

Unai Fischer-Abaigar, Christoph Kern, Frauke Kreuter

Automated decision-making (ADM) systems are being deployed across a diverse range of critical problem areas such as social welfare and healthcare. Recent work highlights the importance of causal ML models in ADM systems, but implementing them in complex social environments poses significant challenges. Research on how these challenges impact the performance in specific downstream decision-making tasks is limited. Addressing this gap, we make use of a comprehensive real-world dataset of jobseekers to illustrate how the performance of a single CATE model can vary significantly across different decision-making scenarios and highlight the differential influence of challenges such as distribution shifts on predictions and allocations.

Unai Fischer-Abaigar, Christoph Kern, Noam Barda et al.

AI-driven decision-making systems are becoming instrumental in the public sector, with applications spanning areas like criminal justice, social welfare, financial fraud detection, and public health. While these systems offer great potential benefits to institutional decision-making processes, such as improved efficiency and reliability, these systems face the challenge of aligning machine learning (ML) models with the complex realities of public sector decision-making. In this paper, we examine five key challenges where misalignment can occur, including distribution shifts, label bias, the influence of past decision-making on the data side, as well as competing objectives and human-in-the-loop on the model output side. Our findings suggest that standard ML methods often rely on assumptions that do not fully account for these complexities, potentially leading to unreliable and harmful predictions. To address this, we propose a shift in modeling efforts from focusing solely on predictive accuracy to improving decision-making outcomes. We offer guidance for selecting appropriate modeling frameworks, including counterfactual prediction and policy learning, by considering how the model estimand connects to the decision-maker's utility. Additionally, we outline technical methods that address specific challenges within each modeling approach. Finally, we argue for the importance of external input from domain experts and stakeholders to ensure that model assumptions and design choices align with real-world policy objectives, taking a step towards harmonizing AI and public sector objectives.

Christoph Kern, Ruben L. Bach, Hannah Mautner et al.

Algorithmic profiling is increasingly used in the public sector as a means to allocate limited public resources effectively and objectively. One example is the prediction-based statistical profiling of job seekers to guide the allocation of support measures by public employment services. However, empirical evaluations of potential side-effects such as unintended discrimination and fairness concerns are rare. In this study, we compare and evaluate statistical models for predicting job seekers' risk of becoming long-term unemployed with respect to prediction performance, fairness metrics, and vulnerabilities to data analysis decisions. Focusing on Germany as a use case, we evaluate profiling models under realistic conditions by utilizing administrative data on job seekers' employment histories that are routinely collected by German public employment services. Besides showing that these data can be used to predict long-term unemployment with competitive levels of accuracy, we highlight that different classification policies have very different fairness implications. We therefore call for rigorous auditing processes before such models are put to practice.

Matthias Kuppler, Christoph Kern, Ruben L. Bach et al.

The advent of powerful prediction algorithms led to increased automation of high-stake decisions regarding the allocation of scarce resources such as government spending and welfare support. This automation bears the risk of perpetuating unwanted discrimination against vulnerable and historically disadvantaged groups. Research on algorithmic discrimination in computer science and other disciplines developed a plethora of fairness metrics to detect and correct discriminatory algorithms. Drawing on robust sociological and philosophical discourse on distributive justice, we identify the limitations and problematic implications of prominent fairness metrics. We show that metrics implementing equality of opportunity only apply when resource allocations are based on deservingness, but fail when allocations should reflect concerns about egalitarianism, sufficiency, and priority. We argue that by cleanly distinguishing between prediction tasks and decision tasks, research on fair machine learning could take better advantage of the rich literature on distributive justice.

Ilja Manakov, Markus Rohm, Christoph Kern et al.

We cast the problem of image denoising as a domain translation problem between high and low noise domains. By modifying the cycleGAN model, we are able to learn a mapping between these domains on unpaired retinal optical coherence tomography images. In quantitative measurements and a qualitative evaluation by ophthalmologists, we show how this approach outperforms other established methods. The results indicate that the network differentiates subtle changes in the level of noise in the image. Further investigation of the model's feature maps reveals that it has learned to distinguish retinal layers and other distinct regions of the images.

Christoph Kern, Bernd Weiss, Jan-Philipp Kolb

Nonresponse in panel studies can lead to a substantial loss in data quality due to its potential to introduce bias and distort survey estimates. Recent work investigates the usage of machine learning to predict nonresponse in advance, such that predicted nonresponse propensities can be used to inform the data collection process. However, predicting nonresponse in panel studies requires accounting for the longitudinal data structure in terms of model building, tuning, and evaluation. This study proposes a longitudinal framework for predicting nonresponse with machine learning and multiple panel waves and illustrates its application. With respect to model building, this approach utilizes information from multiple waves by introducing features that aggregate previous (non)response patterns. Concerning model tuning and evaluation, temporal cross-validation is employed by iterating through pairs of panel waves such that the training and test sets move in time. Implementing this approach with data from a German probability-based mixed-mode panel shows that aggregating information over multiple panel waves can be used to build prediction models with competitive and robust performance over all test waves.