Project descriptions - Page 2
Contact person: Antonio Martini
Status: Project
Prototype to Product: Integrating AI into Complex Systems (P2P AI) focuses on how to successfully move AI from small scale prototypes into everyday use in large organizations and complex IT environments. The project develops practical guidelines and an agile process for integrating AI modules step by step, helping teams balance key factors like maintainability, accuracy, scalability, security, and user experience, and showing when to involve legal, domain, and ethics experts. Starting with real cases at the HISP Center at UiO and extending to TRUST projects and Action Clusters, P2P AI looks for concrete success factors and lessons learned, so users and organizations can reduce risk, avoid common pitfalls, and achieve trustworthy, effective AI integrations in practice.
Contact person: Henrik Litlere Bentsen
Status: Project
How do the public and different stakeholders perceive the implementation and use of Artificial Intelligence across domains? What factors influence fairness/trust perceptions of AI? A goal of the TRUST AI center is to develop trustworthy AI. However, whether or not people trust AI is determined by several factors. This project aims to understand what factors are influencing public perception of and trust in AI systems across different domains and societal groups. To do so, we will leverage the KODEM (Coordinated Online Panels for Research on Democracy and Governance in Norway) national research infrastructure, which provides access to four online panels covering key societal groups (citizens, elected officials, public administrators, and journalists). The KODEM panels are designed to become representative over time and provide high-quality survey data on attitudes and perceptions across the Norwegian population, elected representatives, public administrators, and journalists. Essentially, this project has the potential for relevance across several of the Research Areas in TRUST, because KODEM enables us to investigate what citizens and key societal stakeholders think about AI across a broad range of use cases and domains. The infrastructure also makes it possible to study how these perceptions evolve over time, providing a basis for understanding changing levels of trust and the factors that contribute to trustworthy AI adoption and governance.
Contact Person: Johannes Voll Kolstø
Status: Pre-project
Contact person: Benjamin Ricaud
Status: Project
In this project, we are interested in how the thinking and decision process takes place inside a Large Language Models as it generates text. Understanding it will give us the ability to control, steer and align an LLM with more accuracy and flexibility than doing it during the training phase. We propose to act directly inside the neural network layers to modify the signals and change the model behavior. This direction of research is developing fast but is so far purely empirical with no clear principles. Progress in this direction can lead to better alignment of models, blocking the generation of harmful content, improving reasoning or reducing hallucinations.
Contact person: Runar Hilleren Lie
Status: Project
S2-data is an initiative to establish a research infrastructure for socially sensitive data. It consists of a software platform, a legal infrastructure, and a methodological framework for conducting computationally assisted research on large quantities of data. The software platform allows non-technical researchers to conduct sophisticated computational and AI analysis of large-scale datasets of both structured and unstructured data through a simple interface and without any coding. The project is open source, subject agnostic, and is designed to extend existing university infrastructure. It allows a broad range of research areas to take advantage of the developments in computational research, while ensuring that the research is conducted in a secure and privacy enhanced manner.
Contact person: Martin Jullum
Status: Project
This project addresses an “inference gap” in automated support and filtering for medium- and high-stakes case processing, ranging from customer and citizen services to fraud detection and criminal investigations. In these settings, caseworkers and investigators typically produce case notes and reports explaining why a case receives a particular outcome -- for example, why a complaint is upheld or a claim is judged fraudulent. However, automated systems used to select or prioritize new cases are usually trained only on these outcomes and the structured tabular data available at the time. As a result, the models receive little guidance about the reasoning and information that actually matter in practice. In this project, we aim to develop a Text-Guided Tabular Learning framework, based on LUPI (Learning Using Privileged Information) and cross-modal knowledge distillation, to bridge this gap. A text-aware “teacher” model uses historical reports to guide the training of a regular tabular “student” model, effectively transferring the caseworker’s intuition into an automated system that does not require text when assessing new cases. The results could help users identify relevant cases more accurately, reduce false positives, and make better use of the expert knowledge already captured through case processing, while taking relevant legal and ethical considerations into account.
Contact person: Arnoldo Frigessi
Status: Project
Contact person: Johannes Voll Kolstø
Status: Pre-project.
The pre-project explores robust, generic methods for trend detection in time series: deciding whether a noisy signal is flat or contains a meaningful increase or decrease. The core idea is to recast this as a supervised classification task, training deep learning classifiers on large sets of weakly annotated pairs (X, y), where X is a time series and y a single trend-type label such as "flat" or "contains trend".
This matters because trend detection underlies early event detection across domains, ranging from machine faults to disease outbreaks. Classical change point methods require manual preprocessing and problem-specific modelling, and can perform unsatisfactorily on real, messy data.
The novelty lies in replacing explicit modelling of change and noise processes with a data-centric approach: weak labels make large-scale annotation feasible, capture the context-dependent nature of meaningful trends, and lay the groundwork for future foundation models for trend detection. This pre-project will deliver open-source implementations, pilot studies on data from TRUST partners, and, subject to partner approval, an open benchmark dataset.
Collaboration with user partners is central: the pre-project is motivated by needs from Statkraft and ABB (Action Cluster AC10) and may well be relevant for a variety of TRUST partners. The 6-month pre-project (July 2026 – December 2027) may lead to a larger follow-up project on foundation models for change point detection.
Contact person: Margunn Aanestad
Status: Project
Developing trustworthy AI requires attention to both underlying data, model development and evaluation practices. A central premise for this project is that also implementation – the process of taking the technology into actual use in an organizational context – is an important determinant for trustworthy AI.
We will study real implementation processes with an aim to build knowledge on how responsible implementation can lead to trustworthy AI use. We will do qualitative studies based on observations of work, interviews and document analysis. We focus on how AI is introduced in workflows, whether learning loops, oversight and governance is ensured, and how these actions contribute (or not) to building trust in the technology.
The project team comprises a team at the Department of Informatics (IFI): Margunn Aanestad (contact person), Mari Kannelønning, Alexander Kempton, Troels Mønsted, Miria Grisot, as well as Kristin Asdal (from TIK, SV) and Monika Nerland, Gudrun Rudningen and Lene Pettersen (from IPED, UV).
We will align the project with ongoing research at IFI with Helse Sør-Øst RHF on their large-scale implementation of AI. Given sufficient resources, it would be highly relevant to get access to additional cases across both public and private organizations to study implementation more broadly.
Contact person: Cedric de Coning
Status: Pre-project
AI is increasingly utilised to support international security, conflict prevention, and peacebuilding, where mistrust is a widespread challenge – for instance, among conflict parties and between conflict parties and mediators or peacekeepers. Despite this, little is known about what makes AI applications employed in this area more trustworthy. There is no common understanding of which principles should guide AI use and how to balance priorities such as accuracy, transparency, inclusivity, and explainability. This project aims to fill these gaps by developing a comprehensive model of TrustedAI4Peace&Security, along with practical criteria and measures for responsible AI design and deployment. In collaboration with partners at the UN, we plan to focus on three key areas where trust is critical: remote sensing and object recognition for ceasefire monitoring; earth observation and predictive analytics for early warning of climate-related conflicts; and natural