Manifesto from Dagstuhl Perspectives Workshop 24352 -- Conversational Agents: A Framework for Evaluation (CAFE)

📅 2025-06-08
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
Current conversational information access (CONIAC) systems lack a unified, human-centered evaluation framework. Method: This paper introduces CAFE—the first consensus-driven, multidimensional evaluation framework for CONIAC—systematically integrating six core dimensions: stakeholder goals, user tasks, user characteristics, evaluation criteria, methodologies, and quantitative metrics. Moving beyond traditional technology-centric, unidimensional paradigms, CAFE employs world-model abstraction and interdisciplinary consensus workshops—drawing on human-computer interaction, information retrieval, and evaluation science—to structurally align system objectives with human factors. Contribution/Results: As the first internationally recognized, consensus-driven evaluation framework for conversational agents, CAFE has been formally published in the Dagstuhl Perspectives Workshop series. It provides both theoretical foundations and practical guidelines for the design, evaluation, and standardization of CONIAC systems.

Technology Category

Application Category

📝 Abstract
During the workshop, we deeply discussed what CONversational Information ACcess (CONIAC) is and its unique features, proposing a world model abstracting it, and defined the Conversational Agents Framework for Evaluation (CAFE) for the evaluation of CONIAC systems, consisting of six major components: 1) goals of the system's stakeholders, 2) user tasks to be studied in the evaluation, 3) aspects of the users carrying out the tasks, 4) evaluation criteria to be considered, 5) evaluation methodology to be applied, and 6) measures for the quantitative criteria chosen.
Problem

Research questions and friction points this paper is trying to address.

Define CONversational Information ACcess (CONIAC) and its features
Propose a world model abstracting CONIAC systems
Establish CAFE framework for evaluating CONIAC systems
Innovation

Methods, ideas, or system contributions that make the work stand out.

Proposed a world model for CONIAC systems
Defined CAFE framework with six components
Focused on evaluation criteria and methodology
🔎 Similar Papers
No similar papers found.
Christine Bauer
Christine Bauer
University of Salzburg, AT
L
Li Chen
Hong Kong Baptist University, HK, China
Nicola Ferro
Nicola Ferro
Full Professor, University of Padua, Italy
information retrievaldigital librariesdatabases
Norbert Fuhr
Norbert Fuhr
Univ. Duisburg-Essen
Information Retrieval
Avishek Anand
Avishek Anand
Associate Professor, TU Delft and L3S Research Center
interpretabilitygraph representation learningindexing textranking
Timo Breuer
Timo Breuer
TH Köln - University of Applied Sciences
G
Guglielmo Faggioli
University of Padua, IT
Ophir Frieder
Ophir Frieder
Georgetown University
Hideo Joho
Hideo Joho
Full Professor, Institute of Library, Information and Media Science, University of Tsukuba
Interactive Information RetrievalHuman Information InteractionModel Search Behaviour
Jussi Karlgren
Jussi Karlgren
Stockholm
Language technologyInformation retrievalStylisticsGenreNatural language processing
Johannes Kiesel
Johannes Kiesel
Team lead at GESIS - Leibniz Institute for the Social Sciences
conversational searchcomputational argumentationweb archiving
B
Bart P. Knijnenburg
Clemson University, USA
Aldo Lipani
Aldo Lipani
Professor (Associate) at UCL
AILLMUser SimulationInformation RetrievalNatural Language Processing
Lien Michiels
Lien Michiels
Senior Researcher, Adrem Data Lab University of Antwerp and imec-SMIT, Vrije Universiteit Brussel
evaluationrecommender systemsdiversitynews
A
Andrea Papenmeier
University of Twente, NL
Maria Soledad Pera
Maria Soledad Pera
Associate Professor - Web Information Systems Group - TU Delft
Information RetrievalChildren Information RetrievalInformation AccessRecommender SystemsIR for Non-traditional Users
Mark Sanderson
Mark Sanderson
School of Computing Technologies, RMIT University
Information RetrievalData AnalysisRecommender Systems
Scott Sanner
Scott Sanner
University of Toronto
Artificial IntelligenceMachine LearningInformation Retrieval
Benno Stein
Benno Stein
Professor for Computer Science, Bauhaus-Universität Weimar
Information RetrievalMachine LearningNatural Language ProcessingArtificial Intelligence
Johanne R. Trippas
Johanne R. Trippas
Vice-Chancellor's Senior Research Fellow, School of Computing Technologies, RMIT University
Information RetrievalConversational InterfacesSpoken Interaction DesignConversational Search
Karin Verspoor
Karin Verspoor
Executive Dean, School of Computing Technologies, RMIT University
artificial intelligence in medicinehealth informaticscomputational biologynatural language processingBioNLP
Martijn C. Willemsen
Martijn C. Willemsen
Eindhoven University of Technology & JADS
Decision makingprocess tracingrecommender systemsuser experienceExplainable AI