Project

XAINES

Explaining AI with Narratives

Explaining AI with Narratives

In the XAINES project, the aim is not only to ensure explainability, but also to provide explanations (narratives). The central question is whether AI can explain in one sentence why it acted the way it did or whether it has to explain it interactively to the user. To clarify this, one of the focal points of the project is the exploration of narrative and interactive narratives, which are particularly suitable for humans to absorb knowledge in any form, in their application with AI systems. To obtain explanatory narratives, (linguistically) labelled sensor data streams and predictive models are used. Sensor information is combined with speech information, from which the AI system develops so-called scene understanding, which then generates explanations. Narratives are divided into domain narratives and machine learning narratives. Domain narratives show what happened in the domain as captured by speech-based activity recognition. Machine Learning Narratives are those that explain the predictions of these models. Domain narratives and ML narratives are linked because domain narratives are constructed by machine learning. The end users of these narratives should be, on the one hand, the developers of the AI modules, and on the other hand, the subject matter experts who use the software, but also interested laypersons. The XAINES project, in which 7 DFKI research areas are working closely together, is funded by the Federal Ministry of Education and Research (BMBF). The project follows the new guideline Explainability and Transparency of Machine Learning and Artificial Intelligence (orig.: "Erklärbarkeit und Transparenz des Maschinenllen Lernens und der Künstlichen Intelligenz"), which was launched as part of the German government's AI strategy. The various use cases come from the fields of autonomous driving (ASR), automation in construction (EI) and interactive medical decision support (IML).

Partners

Forschungsbereiche: Agenten und Simulierte Realität (ASR), Interaktives Maschinelles Lernen (IML), Smarte Daten und Wissensdienste (SDS), Eingebettete Intelligenz (EI), Sprachtechnologie (SLT), Sprachtechnologie und Multilingualität (MLT), Algorithmic Business and Production (ABP)

Images

[Translate to English:]

[Translate to English:]

Publications about the project

Rajarshi Biswas, Michael Barz, Mareike Hartmann, Daniel Sonntag

In: Luis Espinosa-Anke , Carlos Martin-Vide , Irena Spasic (editor). Statistical Language and Speech Processing SLSP 2021. International Conference on Statistical Language and Speech Processing (SLSP) 8th-9th November 22-26 Cardiff United Kingdom Lecture Notes in Computer Science / Lecture Notes in Artificial Intelligence (LNCS / LNAI) Springer Heidelberg 11/2021.

To the publication
Rasmus Kær Jørgensen, Xiang Dai, Desmond Elliott, Mareike Hartmann

In: Findings of the Association for Computational Linguistics: EMNLP 2021 Findings of the Association for Computational Linguistics - EMNLP 2021. Conference on Empirical Methods in Natural Language Processing (EMNLP-2021) November 7-11 Online Dominican Republic Pages 3404-3018 1 Association for Computational Linguistics 11/2021.

To the publication
Miryam de Lhoneux, Daniel Hershcovich, Yova Kementchedjhieva, Lukas Nielsen, Chen Qiu, Anders Søgaard, Mareike Hartmann

In: Proceedings of the 25th Conference on Computational Natural Language Learning (CoNLL). Conference on Computational Natural Language Learning (CoNLL-2021) November 10-11 Online Dominican Republic Pages 224-257 Association for Computational Linguistics 11/2021.

To the publication

German Research Center for Artificial Intelligence
Deutsches Forschungszentrum für Künstliche Intelligenz