ELLIS Distinguished Lecture: Hatice Gunes

Item: ELLIS Distinguished Lecture: Hatice Gunes
Rating: 4
Author: PostSingularity Institute

by Hatice Gunes

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQMMachine learning

🎙 Hatice Gunes 👥 3K 📅 February 19, 2026 ⏱ 57 min 👁 87 📄 expert opinion 🧭 2026-08-16

Available in: English (current) Français

Keywords

fairnessbiasaffective computingwellbeingmachine learning

Summary

In this ELLIS Distinguished Lecture, Professor Hatice Gunes from the University of Cambridge presents her research on fairness for affective and wellbeing computing systems and agents. She begins by highlighting the importance of addressing AI bias, referencing the pioneering work of Buolamwini and Gebru on gender classification. She defines key concepts such as sensitive attributes, dataset bias, and algorithmic bias, and introduces a framework for bias analysis and mitigation at three levels: pre-processing, in-processing, and post-processing. Using the RAFD dataset as an example, she demonstrates how data augmentation, fairness through awareness, and disentangled approaches can improve fairness metrics. She then discusses applications in mental health disorder prediction, emphasizing the need for fairness in high-stakes scenarios. The talk concludes with recommendations for future research, including the need for context-sensitive interpretation and the development of fairer AI systems.

136 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a comprehensive overview of fairness in affective computing, grounded in the speaker’s extensive research. The argumentation is solid, as she systematically explains bias types, evaluation metrics, and mitigation techniques, supported by concrete examples and results from her own studies. She effectively communicates the trade-offs between accuracy and fairness and highlights the importance of considering multiple fairness metrics. The presentation is well-structured and accessible, making a strong case for integrating fairness considerations throughout the AI development pipeline.

Scientific Rigor, Source Quality, Title Accuracy

The talk demonstrates scientific rigor through the use of published research and established datasets. The speaker references her own work, including the ‘Hitchhiker’s Guide to Bias and Fairness in Facial Affective Signal Processing’ and recent papers on continual learning and optimizers. The title accurately reflects the content, as the talk focuses on fairness in affective and wellbeing computing systems. The sources cited are credible and relevant, though the talk does not provide external references beyond the speaker’s own publications and the ELLIS Institute link.

178 words

Title / Content Match

The title accurately reflects the content, as the talk focuses on fairness in affective and wellbeing computing systems and agents.

Quality & Reliability

8/10

The talk is given by a leading researcher in affective computing, with a strong publication record and recognition. The content is based on her own research and published works, and she provides concrete examples and references to her own papers. However, the talk is a lecture, not a peer-reviewed publication, and some claims are not fully detailed.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

ELLIS Distinguished Lecture: Hatice Gunes — Official page for the lecture, providing context and possibly additional resources.

Concurring Sources

Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification — The speaker references this work as a pioneering study highlighting bias in gender classification systems.

Contribution & Novelties

The talk synthesizes the speaker’s research on fairness in affective computing, offering a practical guide for bias analysis and mitigation. It emphasizes the importance of considering fairness in high-stakes applications like mental health prediction and provides concrete examples from her work. The talk also highlights the need for context-sensitive interpretation in affective computing, which is a novel perspective.

Pour aller plus loin :

Buolamwini and Gebru (2018) - Gender Shades — This seminal paper exposed bias in commercial gender classification systems, foundational to the talk’s motivation.
Hitchhiker’s Guide to Bias and Fairness in Facial Affective Signal Processing — The speaker’s own guide, providing a comprehensive overview of bias types and mitigation methods.
Fairness in Machine Learning (Barocas et al.) — An authoritative resource on fairness definitions and algorithms, relevant for deeper understanding.

132 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, as well as technical level, indicating a dense and well-structured presentation. The global reliability is also high, reflecting the speaker's expertise and use of published research. The overall score is strong, with minor deductions for the lack of external sources and the lecture format.

Reliability 8/10