Algorithms with People

New Behavioral Paradigms for Characterizing Representations

As large-scale online experiments are becoming increasingly configurable and adaptive, they blur the line between behavioral paradigms and computer algorithms. This perspective (sometimes referred to as Algorithms with People) inspires a new class of paradigms for characterizing representations. By treating representations as subjective probabilities we can design behavioral experiments that instantiate sampling algorithms from those representations. I have been involved in multiple projects that develop and leverage such paradigms for studying semantic and perceptual representations in humans and machines, including (i) Gibbs Sampling with People (Harrison* et al., 2020; Marjieh et al., 2024; Kumar et al., 2022), (ii) STEP-Tag (Marjieh et al., 2023; Marjieh et al., 2024), and (iii) Multimodal serial reproduction (Marjieh et al., 2023; Kumar* et al., 2024). For a recent summary of this approach, check out this book chapter that we recently wrote (Griffiths et al., 2024).

References

2024

  1. consonance.png
    Timbral effects on consonance disentangle psychoacoustic mechanisms and suggest perceptual origins for musical scales
    Raja Marjieh, Peter MC Harrison, Harin Lee, and 2 more authors
    Nature Communications. Listen to the experiments here , 2024
  2. groundedsem.png
    Characterizing the Large-Scale Structure of Grounded Semantic Networks
    Raja Marjieh, Pol Rijn, Ilia Sucholutsky, and 3 more authors
    PsyArXiv, 2024
  3. multimodalsr.png
    Comparing Abstraction in Humans and Machines Using Multimodal Serial Reproduction
    Sreejan Kumar*Raja Marjieh*, Byron Zhang, and 5 more authors
    In Proceedings of the Annual Meeting of the Cognitive Science Society, 2024
  4. samplingsubj.png
    Estimating Subjective Probability Distributions
    Thomas L Griffiths, Adam N Sanborn, Raja Marjieh, and 3 more authors
    In Bayesian Models of Cognition: Reverse Engineering the Mind, 2024

2023

  1. words.png
    Words are all you need? Language as an approximation for human similarity judgments
    Raja Marjieh, Pol Van Rijn, Ilia Sucholutsky, and 4 more authors
    In The Eleventh International Conference on Learning Representations, 2023
  2. diffusionsr.gif
    Analyzing Diffusion as Serial Reproduction
    Raja Marjieh, Ilia Sucholutsky, Thomas A Langlois, and 2 more authors
    In Proceedings of the 40th International Conference on Machine Learning, 2023

2022

  1. gridgsp.png
    Using natural language and program abstractions to instill human inductive biases in machines
    Sreejan Kumar, Carlos G Correa, Ishita Dasgupta, and 7 more authors
    Advances in Neural Information Processing Systems. Outstanding Paper Award , 2022

2020

  1. colorgsp.png
    Gibbs Sampling with People
    Peter Harrison*Raja Marjieh*, Federico Adolfi, and 5 more authors
    In Advances in Neural Information Processing Systems. Oral Presentation , 2020