Tom Sühr

Tom Sühr, Florian E. Dorner, Samira Samadi et al.

With large language models (LLMs) like GPT-4 appearing to behave increasingly human-like in text-based interactions, it has become popular to attempt to evaluate personality traits of LLMs using questionnaires originally developed for humans. While reusing measures is a resource-efficient way to evaluate LLMs, careful adaptations are usually required to ensure that assessment results are valid even across human subpopulations. In this work, we provide evidence that LLMs' responses to personality tests systematically deviate from human responses, implying that the results of these tests cannot be interpreted in the same way. Concretely, reverse-coded items ("I am introverted" vs. "I am extraverted") are often both answered affirmatively. Furthermore, variation across prompts designed to "steer" LLMs to simulate particular personality types does not follow the clear separation into five independent personality factors from human samples. In light of these results, we believe that it is important to investigate tests' validity for LLMs before drawing strong conclusions about potentially ill-defined concepts like LLMs' "personality".

Mirabel Reid, Tom Sühr, Claire Vernade et al.

Machine Learning (ML) models are increasingly used to support or substitute decision making. In applications where skilled experts are a limited resource, it is crucial to reduce their burden and automate decisions when the performance of an ML model is at least of equal quality. However, models are often pre-trained and fixed, while tasks arrive sequentially and their distribution may shift. In that case, the respective performance of the decision makers may change, and the deferral algorithm must remain adaptive. We propose a contextual bandit model of this online decision making problem. Our framework includes budget constraints and different types of partial feedback models. Beyond the theoretical guarantees of our algorithm, we propose efficient extensions that achieve remarkable performance on real-world datasets.

Meike Zehlike, Tom Sühr, Carlos Castillo et al.

Ranked search results and recommendations have become the main mechanism by which we find content, products, places, and people online. With hiring, selecting, purchasing, and dating being increasingly mediated by algorithms, rankings may determine career and business opportunities, educational placement, access to benefits, and even social and reproductive success. It is therefore of societal and ethical importance to ask whether search results can demote, marginalize, or exclude individuals of unprivileged groups or promote products with undesired features. In this paper we present FairSearch, the first fair open source search API to provide fairness notions in ranked search results. We implement two algorithms from the fair ranking literature, namely FA*IR (Zehlike et al., 2017) and DELTR (Zehlike and Castillo, 2018) and provide them as stand-alone libraries in Python and Java. Additionally we implement interfaces to Elasticsearch for both algorithms, that use the aforementioned Java libraries and are then provided as Elasticsearch plugins. Elasticsearch is a well-known search engine API based on Apache Lucene. With our plugins we enable search engine developers who wish to ensure fair search results of different styles to easily integrate DELTR and FA*IR into their existing Elasticsearch environment.

Tom Sühr, Florian E. Dorner, Olawale Salaudeen et al.

Large Language Models (LLMs) have achieved remarkable results on a range of standardized tests originally designed to assess human cognitive and psychological traits, such as intelligence and personality. While these results are often interpreted as strong evidence of human-like characteristics in LLMs, this paper argues that such interpretations constitute an ontological error. Human psychological and educational tests are theory-driven measurement instruments, calibrated to a specific human population. Applying these tests to non-human subjects without empirical validation, risks mischaracterizing what is being measured. Furthermore, a growing trend frames AI performance on benchmarks as measurements of traits such as ``intelligence'', despite known issues with validity, data contamination, cultural bias and sensitivity to superficial prompt changes. We argue that interpreting benchmark performance as measurements of human-like traits, lacks sufficient theoretical and empirical justification. This leads to our position: Stop Evaluating AI with Human Tests, Develop Principled, AI-specific Tests instead. We call for the development of principled, AI-specific evaluation frameworks tailored to AI systems. Such frameworks might build on existing frameworks for constructing and validating psychometrics tests, or could be created entirely from scratch to fit the unique context of AI.

Tom Sühr, Samira Samadi, Chiara Farronato

Machine learning (ML) models are increasingly used in various applications, from recommendation systems in e-commerce to diagnosis prediction in healthcare. In this paper, we present a novel dynamic framework for thinking about the deployment of ML models in a performative, human-ML collaborative system. In our framework, the introduction of ML recommendations changes the data-generating process of human decisions, which are only a proxy to the ground truth and which are then used to train future versions of the model. We show that this dynamic process in principle can converge to different stable points, i.e. where the ML model and the Human+ML system have the same performance. Some of these stable points are suboptimal with respect to the actual ground truth. As a proof of concept, we conduct an empirical user study with 1,408 participants. In the study, humans solve instances of the knapsack problem with the help of machine learning predictions of varying performance. This is an ideal setting because we can identify the actual ground truth, and evaluate the performance of human decisions supported by ML recommendations. We find that for many levels of ML performance, humans can improve upon the ML predictions. We also find that the improvement could be even higher if humans rationally followed the ML recommendations. Finally, we test whether monetary incentives can increase the quality of human decisions, but we fail to find any positive effect. Using our empirical data to approximate our collaborative system suggests that the learning process would dynamically reach an equilibrium performance that is around 92% of the maximum knapsack value. Our results have practical implications for the deployment of ML models in contexts where human decisions may deviate from the indisputable ground truth.

Meike Zehlike, Tom Sühr, Carlos Castillo

In this report we provide an improvement of the significance adjustment from the FA*IR algorithm of Zehlike et al., which did not work for very short rankings in combination with a low minimum proportion $p$ for the protected group. We show how the minimum number of protected candidates per ranking position can be calculated exactly and provide a mapping from the continuous space of significance levels ($α$) to a discrete space of tables, which allows us to find $α_c$ using a binary search heuristic.

Tom Sühr, Sophie Hilgard, Himabindu Lakkaraju

Ranking algorithms are being widely employed in various online hiring platforms including LinkedIn, TaskRabbit, and Fiverr. Prior research has demonstrated that ranking algorithms employed by these platforms are prone to a variety of undesirable biases, leading to the proposal of fair ranking algorithms (e.g., Det-Greedy) which increase exposure of underrepresented candidates. However, there is little to no work that explores whether fair ranking algorithms actually improve real world outcomes (e.g., hiring decisions) for underrepresented groups. Furthermore, there is no clear understanding as to how other factors (e.g., job context, inherent biases of the employers) may impact the efficacy of fair ranking in practice. In this work, we analyze various sources of gender biases in online hiring platforms, including the job context and inherent biases of employers and establish how these factors interact with ranking algorithms to affect hiring decisions. To the best of our knowledge, this work makes the first attempt at studying the interplay between the aforementioned factors in the context of online hiring. We carry out a largescale user study simulating online hiring scenarios with data from TaskRabbit, a popular online freelancing site. Our results demonstrate that while fair ranking algorithms generally improve the selection rates of underrepresented minorities, their effectiveness relies heavily on the job contexts and candidate profiles.