Iyad Rahwan

Max Planck Society
2 previous affiliations
15Machine learning papers tracked by Lacuna
Is this you?

Research Overview

Iyad Rahwan focuses on computational social science, artificial intelligence evaluation, and the ethical dimensions of automated systems. His trajectory spans from leveraging large-scale social media data for sentiment analysis to developing frameworks for algorithmic auditing and societal preference aggregation. Most recently, his work investigates open-ended art generation, exploring methods to balance originality and coherence in creative artificial intelligence.

Research Trajectory

Early in the observed period, Rahwan contributed to natural language processing research addressing the scarcity of manually annotated data. In the 2017 paper Using millions of emoji occurrences to learn any-domain representations for detecting sentiment, emotion and sarcasm, he and his collaborators demonstrated that extending distant supervision to a diverse set of noisy labels—specifically, millions of emoji occurrences—enables models to learn richer representations. This work achieved strong performance across benchmark datasets for sentiment, emotion, and sarcasm detection using a single pretrained model, contributing to the broader area of Distant Supervision for Scaling Emotion Recognition as well as the research direction concerning The Temporal Decay of Social Media Datasets.

The following year, his collaborative efforts expanded into the intersection of machine learning and computational social choice. In A Voting-Based System for Ethical Decision Making, the research team proposed a general approach to automating ethical decisions by learning a model of societal preferences. When faced with an ethical dilemma at runtime, the system aggregates these preferences to identify a desirable choice. The researchers implemented and evaluated this system in the domain of autonomous vehicles, utilizing preference data collected from over a million participants through the Moral Machine project. Concurrently, Rahwan directed work on democratizing the scientific study of artificial intelligence. He oversaw the development of TuringBox: An Experimental Platform for the Evaluation of AI Systems, an architecture designed to facilitate algorithmic auditing. The platform allows AI contributors to upload algorithms for study, while examiners can develop and post machine intelligence tasks to evaluate and characterize the algorithmic outputs.

Returning to the dynamics of social media platforms, Rahwan participated in a 2023 study examining the algorithmic optimization of user engagement and its relationship to echo chambers. The paper Optimal Engagement-Diversity Tradeoffs in Social Media introduced a novel model to question whether echo chambers are an inevitable consequence of high engagement. The research established theoretical bounds on the maximum engagement achievable under a diversity constraint, quantifying the worst-case tradeoff between these two objectives, and charted the Pareto frontier using empirical data from Twitter.

Most recently, Rahwan has guided research toward generative models and computational creativity. In 2025, his group introduced the Cultural Alien Sampler: Open-ended Art Generation Balancing Originality and Coherence. This method explicitly separates compositional fit from cultural typicality to generate novel and harmonious art. By demonstrating that strategically deviating from cultural norms can unlock creative potential in AI, this work advances the research direction of Balancing Novelty and Coherence in Creative AI.

Current Directions

Recent work investigates the intersection of artificial intelligence and computational creativity, with a specific focus on open-ended art generation. The most recent papers in the evidence address the challenge of designing generative systems that can produce outputs that are both highly original and structurally coherent. This involves exploring methods to decouple compositional harmony from cultural typicality, allowing AI systems to strategically deviate from established norms to discover novel creative spaces. Additionally, ongoing themes in the broader research arc continue to touch upon the societal impacts of algorithmic systems, including how platforms balance user engagement with content diversity and how human preferences can be aggregated to inform the behavior of autonomous systems.

Profile updated

See something wrong? Report a mistake