06 October 2023
Artificial Intelligence, and in particular Machine Learning (ML), is nowadays ubiquitous thanks to massive investments in making it a commodity. This is accompanied by increasing concerns about AI's impact on society. In fact, since real-world datasets often reflect historical biases in society, when they are fed to ML algorithms, it often results in models that exacerbate these biases.
In this work, EMERGE partners from the University of Pisa address the problem of algorithmic fairness in Machine Learning for temporal data, focusing on ensuring that sensitive time-dependent information does not unfairly influence the outcome of a classifier.
In particular, the authors focus on a class of training-efficient recurrent neural models called Echo State Networks, and show, for the first time, how to leverage local unsupervised adaptation of the internal dynamics to build fairer classifiers. Experimental results on real-world problems from physiological sensor data demonstrate the potential of the proposal.
Read the paper: https://doi.org/10.14428/esann/2023.ES2023-90
More Information
