UPF invites applications for a fellowship focused on advancing AI applications in Cultural Heritage and Inclusive Societies. The selected candidate(s) will work across diverse topics, including developing AI tools for accessibility, improving fairness in decision-making systems, and enhancing the accessibility of digital content. Projects in this cluster aim to address pressing societal challenges by creating ethical, interpretable, and inclusive AI systems.
Key areas of focus include*:
- Natural Language Processing for Accessibility: Design advanced text simplification techniques combining neural models and human-interpretable specifications to create inclusive content for individuals with diverse reading, writing, and comprehension abilities.
- Digital Media Accessibility: Develop AI-based multimodal solutions, such as video and image restoration, sign language translation, and audio-visual content enhancement, to make media accessible to individuals with visual, hearing, or reading impairments.
- Scientific Knowledge Simplification: Create tools for the aggregation, summarization, and simplification of scientific documents to improve accessibility for researchers, educators, and citizens in the context of open science.
- Ethical AI Systems: Research fairness, accountability, transparency, and non-discrimination in high-risk decision-support systems, particularly those impacting access to essential services, criminal justice, and education.
- Algorithmic Auditing: Develop methodologies for assessing and mitigating risks of bias in AI systems, including discrimination risk assessments and post-deployment monitoring for equity.
Key Responsibilities*:
- Design and implement advanced text simplification techniques that combine neural models and human-interpretable specifications, enabling inclusive content creation for individuals with diverse accessibility needs.
- Develop multimodal AI solutions for enhancing video, audio, and image content accessibility, including sign language translation, video restoration, and sound localization, to support individuals with visual, hearing, or reading impairments.
- Create tools for the aggregation, summarization, and simplification of scientific documents, facilitating better access to interconnected knowledge for researchers, educators, and citizens in the open science context.
- Research and design fairness-driven decision-support systems for high-stakes applications, focusing on algorithmic transparency, accountability, and equitable treatment across diverse populations.
- Develop and apply methodologies for assessing bias and discrimination risks in AI systems, proposing innovative strategies for ongoing monitoring and ensuring equity post-deployment.
*Candidates are not expected to focus on or address all these items simultaneously; these represent the broader research directions of the role.
Qualifications:
- PhD in Artificial Intelligence, Social Computing, Computer Science, or related fields.
- Expertise in NLP, multimodal learning, or accessibility-focused AI applications.
- Strong background in algorithmic fairness, responsible AI, or transparency in decision systems.
- Excellent problem-solving and collaboration skills, with a commitment to ethical AI practices.
Available Infrastructures: UPF houses infrastructure for both engineering work (computing, robotics and sensing, specialised audio-visual equipment – VR, audio, etc.) and experimentation with human subjects (experimental rooms with specialised equipment), complemented with links to hospitals and other external facilities.
Possible Principal Investigators:
- Horacio Saggion leads research on natural language technologies for access to information in inclusive societies and on intelligent access to multimodal document collections.
Email: horacio.saggion@upf.edu
Web: https://www.upf.edu/web/horacio-saggion
Possible Secondments: Partners in the Horizon Europe project iDEM (https://idemproject.eu/); Industrial Partners currently collaborating with the principal investigator in scientific text mining.
Keywords: Large Language Models; Information Extraction; Text Generation; Text Simplification; Text Summarization; Multimodality; Multi-linguality; Inclusive Societies; Accessibility.
- Coloma Ballester, Pablo Arias, Gloria Haro, and Federico Sukno lead research on digital content enhancement to increase the accessibility of people with visual, hearing or reading impairment to available video content.
Email: coloma.ballester@upf.edu
Web: https://www.upf.edu/web/ipcv
Possible Secondments: ENS Paris-Saclay
Web: https://www.upf.edu/web/ipcv
Possible Secondments: ENS Paris-Saclay
Keywords: Computer Vision; Multimodal Learning; Generative Models; Human Body/Face, Pose, and Gesture; Low-Level Vision.
- Carlos Castillo leads research on fairness, accountability, and transparency in decision support systems powered by AI, particularly those in high-risk areas of AI, as well as on the development and evaluation of non-discriminatory AI systems.
Email: carlos.castillo@upf.edu
Web: https://www.chato.cl/research/discrimination_bias
Possible Secondments: Telefónica, Adevinta, Eticas Research and Consulting.
Keywords: Algorithmic Fairness; Algorithmic Discrimination; Algorithmic Auditing; Transparency; Explainable AI; Responsible AI; Trustworthy AI.
