AISL — Accountable Intelligent Systems Lab

Collaboration

Research Collaboration

AISL welcomes research and innovation collaborations with industry, public institutions, and academic partners. We are open to technology consulting, AI strategy consulting, applied research, commissioned research, and collaborative research projects.

We actively participate in national and international research initiatives, including projects funded by Innosuisse, the Swiss National Science Foundation (SNSF), Horizon Europe, and other competitive funding programs.

Start a Collaboration

AI Strategy & Consulting

Hybrid AI & LLMs
Agent Systems
AI transformation
Human-centered AI

Applied Research

Contract research
Prototype development
Feasibility studies
Technology transfer

Research Partnerships

Innosuisse
SNSF
Horizon Europe
International collaborations

Lab Mission

Accountable AI for sensitive and safety-critical environments

AISL researches hybrid, neuro- and intersymbolic AI architectures that combine machine learning, generative models, and symbolic knowledge. The aim is to make intelligent systems more transparent, explainable, verifiable, and controllable when they are used in regulated or high-stakes settings.

In these environments, probabilistic behavior alone is not enough. Our work explores deterministic guardrails for protecting data privacy and the emotional safety of vulnerable people, and for supporting the reliability of critical infrastructure, organizations, and public services.

Research

Research focus

AISL works on AI systems that connect language, knowledge, reasoning, and action while keeping privacy, safety, accountability, and robustness at the center of the design.

Research Pillars

Conversational and Dialogue AI

Hybrid dialogue systems, conversational agents, and human-AI interaction for sensitive settings such as healthcare, public services, education, and organizational decision support.

Agentic AI and LxM Agents

Agent architectures that combine language understanding with reasoning, planning, orchestration, and action across digital services and software systems.

Neuro- and Intersymbolic AI

Architectures that integrate machine learning with knowledge graphs, ontologies, symbolic representations, and logical reasoning.

Research Topics

Dialogue Systems, Generative & Agentic AI

We study conversational AI systems that combine classical NLP, large language models, and knowledge-based reasoning. The focus is on natural interaction without losing reliability, traceability, or control.

Hybrid Dialogue Systems

Our hybrid dialogue architectures combine intent-based dialogue management with generative models and knowledge-grounded reasoning. They can shift between structured control and generative responses depending on task, context, and interaction risk.

Responsible Use of Generative AI

We investigate verification and control mechanisms for generative AI, including retrieval-augmented generation, semantic validation with knowledge graphs, and hybrid reasoning approaches that reduce misinformation, privacy, and safety risks.

Alignment and Domain Adaptation

The lab studies how language models can be aligned and adapted to specific tasks and sensitive domains, using fine-tuning, preference alignment, reinforcement-based capability alignment, and neuro- and intersymbolic safeguards.

Projects

Selected research projects

Our projects translate research on dialogue systems, healthcare coaching, legal AI, organizational intelligence, and knowledge-based reasoning into applied prototypes and partner-driven studies.

Innosuisse

RepoChat

Knowledge-grounded dialogue system that combines LLMs, retrieval mechanisms, and knowledge graphs to provide verifiable access to complex technical documents.

Industry-funded

RepoChat Go-Public

Scaling and quality assurance of a knowledge-grounded dialogue system combining LLMs, retrieval mechanisms, and knowledge graphs for verifiable access to extensive technical documentation.

Consulting

Trustworthy AI Procurement

Vendor-neutral assessment of generative AI, language models, and agent architectures for a complex public-sector platform procurement.

SNSF SPIRIT

Healthcare Coaches

Intelligent health chatbot research for behavioral change and improved health outcomes in sensitive public health contexts.

Industry-funded

Long-Context LLMs

Hybrid approach combining long-context language models, human expertise, and clustering techniques to detect inconsistencies in large document collections.

Eurostars / Innosuisse

LeadAI

AI-driven organizational digital twins and virtual assistants for leadership consulting, organizational analysis, and decision support.

FHNW

ChEdventure

Chatbot-based educational simulation for developing questioning, critical thinking, workshop facilitation, and contradiction-resolution skills.

FHNW

Digital Self-Study Assistant

Intelligent tutoring system supporting coaching, reflection, and learning progress in project-based education.

Innosuisse

Legal Chatbot Workbench

Platform concept enabling organizations such as law firms and professional associations to develop legal chatbots with minimal technical effort.

Innosuisse

Ontology-Based Case-Based Reasoning

Ontology-based reasoning mechanisms and enterprise knowledge models for supporting complex software integration projects.

Teaching

Teaching and talent development

AISL connects research with teaching by bringing current work on dialogue systems, hybrid AI, responsible generative AI, and AI-assisted software development into the classroom.

Generative AI & Agent Systems (BSc BAI)

LxM architectures, large language models, pre-training, mid-training, post-training, alignment, and domain-specific foundation models.

Dialogue Systems and Natural Language Processing (BSc BAI)

Conversational design, dialogue management, chatbot engineering, retrieval-augmented generation, and computational linguistics.

AI-Assisted Software Development (BSc BAI / BIT)

Lean software development, test-driven development, AI-assisted coding, coding agents, and agentic software engineering.

AI Operations (BSc BAI)

Model deployment, MLOps, API provisioning, monitoring, governance, lifecycle management, and scaling AI systems in production environments.

AI for Business Processes (MSc BIS / MI)

Intelligent process support, hyperautomation, process mining, agent orchestration, and conversational business systems.

Community

Research community and editorial service

AISL contributes to communities concerned with hybrid, trustworthy, applied, and knowledge-grounded AI.

MAKE Symposium Series

The MAKE symposium series is an international forum for the principled integration of machine learning and knowledge engineering within the AAAI Spring Symposium Series. Prof. Martin serves as founding organizer, chair, and organizing and program committee member across multiple editions focused on hybrid, trustworthy, and knowledge-grounded AI.

Visit MAKE Symposium Series

Applied Artificial Intelligence Journal

Prof. Martin serves as Associate Editor of the Taylor & Francis journal Applied Artificial Intelligence, which publishes peer-reviewed research on AI applications in management, industry, engineering, administration, and education.

Visit Applied Artificial Intelligence

arXiv cs.AI Moderation

The moderators of arXiv cs.AI (Computer Science – Artificial Intelligence), including Prof. Martin, review incoming submissions for topical fit, scientific quality, and adherence to arXiv's endorsement and content policies, distinguishing AI-focused work from adjacent subareas.

Visit arXiv cs.AI

People