Prompt-based language models like GPT4 and LLaMa have been used for a wide variety of use cases such as simulating agents, searching for information, or for content analysis. For all of these applications and others, political biases in these models can affect their performance. Several researchers have attempted to study political bias in language models using evaluation suites based on surveys, such as the Political Compass Test (PCT), often finding a particular leaning favored by these models. However, there is some variation in the exact prompting techniques, leading to diverging findings and most research relies on constrained-answer settings to extract model responses. Moreover, the Political Compass Test is not a scientifically valid survey instrument. In this work, we contribute a political bias measured informed by political science theory, building on survey design principles to test a wide variety of input prompts, while taking into account prompt sensitivity. We then prompt 11 different open and commercial models, differentiating between instruction-tuned and non-instruction-tuned models, and automatically classify their political stances from 88,110 responses. Leveraging this dataset, we compute political bias profiles across different prompt variations and find that while PCT exaggerates bias in certain models like GPT3.5, measures of political bias are often unstable, but generally more left-leaning for instruction-tuned models.

Generative language models are increasingly being subjected to psychometric questionnaires intended for human testing, in efforts to establish their traits, as benchmarks for alignment, or to simulate participants in social science experiments. While this growing body of work sheds light on the likeness of model responses to those of humans, concerns are warranted regarding the rigour and reproducibility with which these experiments may be conducted. Instabilities in model outputs, sensitivity to prompt design, parameter settings, and a large number of available model versions increase documentation requirements. Consequently, generalization of findings is often complex and reproducibility is far from guaranteed. In this paper, we present R.U.Psycho, a framework for designing and running robust and reproducible psychometric experiments on generative language models that requires limited coding expertise. We demonstrate the capability of our framework on a variety of psychometric questionnaires, which lend support to prior findings in the literature. R.U.Psycho is available as a Python package at https://github.com/julianschelb/rupsycho.

Writing assistants and large language models see widespread use in the creation of text content. While their effectiveness for individual users has been evaluated in the literature, little is known about their proclivity to change language or reduce its richness when adopted by a large user base. In this paper, we take a first step towards quantifying this risk by measuring the semantic and vocabulary change enacted by the use of rephrasing tools on a multi-domain corpus of human-generated text.

The growing popularity of generative language models has amplified interest in interactive methods to guide model outputs. Prompt refinement is considered one of the most effective means to influence output among these methods. We identify several challenges associated with prompting large language models, categorized into data- and model-specific, linguistic, and socio-linguistic challenges. A comprehensive examination of model outputs, including runner-up candidates and their corresponding probabilities, is needed to address these issues. The beam search tree, the prevalent algorithm to sample model outputs, can inherently supply this information. Consequently, we introduce an interactive visual method for investigating the beam search tree, facilitating analysis of the decisions made by the model during generation. We quantitatively show the value of exposing the beam search tree and present five detailed analysis scenarios addressing the identified challenges. Our methodology validates existing results and offers additional insights.

The awareness and mitigation of biases are of fundamental importance for the fair and transparent use of contextual language models, yet they crucially depend on the accurate detection of biases as a precursor. Consequently, numerous bias detection methods have been proposed, which vary in their approach, the considered type of bias, and the data used for evaluation. However, while most detection methods are derived from the word embedding association test for static word embeddings, the reported results are heterogeneous, inconsistent, and ultimately inconclusive. To address this issue, we conduct a rigorous analysis and comparison of bias detection methods for contextual language models. Our results show that minor design and implementation decisions (or errors) have a substantial and often significant impact on the derived bias scores. Overall, we find the state of the field to be both worse than previously acknowledged due to systematic and propagated errors in implementations, yet better than anticipated since divergent results in the literature homogenize after accounting for implementation errors. Based on our findings, we conclude with a discussion of paths towards more robust and consistent bias detection methods.

In real-world applications of influence maximization (IM), the network structure is often unknown. In this case, we may identify the most influential seed nodes by exploring only a part of the underlying network given a small budget for node queries. Motivated by the fact that collecting node metadata is more cost-effective than investigating the relationship between nodes via queried nodes, we develop IM-META, an end-to-end solution to IM in networks with unknown topology by retrieving information from both queries and node metadata. However, using such metadata to aid the IM process is not without risk due to the noisy nature of metadata and uncertainties in connectivity inference. To tackle these challenges, we formulate an IM problem that aims to find two sets, i.e., seed nodes and queried nodes. We propose an effective method that iteratively performs three steps: 1) we learn the relationship between collected metadata and edges via a Siamese neural network model, 2) we select a number of inferred influential edges to construct a reinforced graph used for discovering an optimal seed set, and 3) we identify the next node to query by maximizing the inferred influence spread using a topology-aware ranking strategy. By querying only 5% of nodes, IM-META reaches 93% of the upper bound performance.

Wikipedia is one of the most popular sources of free data on the Internet and subject to extensive use in numerous areas of research. Wikidata on the other hand, the knowledge base behind Wikipedia, is less popular as a source of data, despite having the "data" already in its name, and despite the fact that many applications in Natural Language Processing in general and Information Extraction in particular benefit immensely from the integration of knowledge bases. In part, this imbalance is owed to the younger age of Wikidata, which launched over a decade after Wikipedia. However, this is also owed to challenges posed by the still evolving properties of Wikidata that make its content more difficult to consume for third parties than is desirable. In this article, we analzye the causes of these challenges from the viewpoint of a data consumer and discuss possible avenues of research and advancement that both the scientific and the Wikidata community can collaborate on to turn the knowledge base into the invaluable asset that it is uniquely positioned to become.