Trustworthy access to legal information

A terminated rental contract, an unexpected bill, a dispute at work: legal questions often arise suddenly, and understanding what the law says can be difficult without professional support. Searching online may provide quick answers, but not necessarily reliable ones. Consulting a lawyer, meanwhile, can be costly. For Niklas Traser, this gap is a research problem.

As a new PhD researcher in Ivan Habernal’s Trustworthy Human Language Technologies (TrustHLT) group, Niklas Traser is working on TruQALIA – “Trustworthy Question Answering to Enable Legal Information Access for Laypeople.” The project explores how natural language processing can help people find and understand trustworthy legal information for everyday questions. TruQALIA is a collaborative project with the University of Trier, where a legal researcher will examine the legal implications of the approaches developed in the project.

The emphasis is important: the goal is not simply to make language models answer more legal questions. It is to investigate how such systems can provide useful information while their limitations and risks remain visible and understood.

Trust before convenience

Language models can make complex information remarkably easy to access. But legal contexts leave little room for confident-sounding mistakes. For Niklas Traser, this is precisely what makes the field compelling. He sees the potential to make legal information more accessible to people who may not easily be able to seek professional legal support. At the same time, he wants to understand where current language technologies fail and how they can become more reliable.

That combination of accessibility and caution places his research directly within the broader questions pursued by TrustHLT. At the Faculty of Computer Science at Ruhr University Bochum, Habernal’s group conducts research in privacy-preserving NLP and legal NLP, examining how language technologies can be made trustworthy in sensitive domains.

As part of the Research Center Trustworthy Data Science and Security (RC Trust), the group also connects these questions to a wider interdisciplinary challenge: AI systems should not only be powerful enough to perform a task. People need to know when their results can be relied upon – and when they cannot.

From fact-checking to Legal NLP

Traser brings a strong NLP background to Bochum. He completed both his Bachelor’s and Master’s degrees in Computer Science at TU Darmstadt, increasingly focusing on natural language processing during his graduate studies.

At the university’s UKP Lab, he worked as a student assistant on research related to multimodal fact-checking and later completed his Master’s thesis there. These experiences shaped an interest in NLP systems that interact with complex real-world information rather than operating only under ideal laboratory conditions. That made Ivan Habernal’s group a natural next step.

For Traser, TrustHLT brings together several strands that have interested him throughout his studies: NLP, trustworthy AI, and applications with tangible real-world relevance. He is particularly interested in the group’s focus on understanding the limitations of current systems rather than treating technical performance alone as sufficient.

Making legal AI useful in practice

During his PhD, Niklas Traser hopes to develop methods that are scientifically interesting but also useful beyond research settings. TruQALIA gives him a particularly demanding testing ground: making legal information easier to access while taking the limitations and risks of language models seriously.

It is a challenge that reaches far beyond building a better question-answering system. If AI is to help people navigate areas as consequential as law, accessibility and trustworthiness have to develop together.

Outside research, Niklas Traser switches between somewhat different kinds of instruments: he plays guitar, piano, and trumpet. He also enjoys biking and bouldering – useful counterpoints to the questions of language models, law, and trustworthy AI that now shape his work in Bochum.