Auditing Algorithmic Bias in Transformer-Based Trading

Armin Gerami, Ramani Duraiswami

Keywords: Transformers, Trading, Partial Information Decomposition, Bias

TL;DR: We audited a transformer model used for financial predictions and found two concerning biases: the model completely ignores data volatility when making decisions, and it favors assets with less frequent price changes over more volatile ones.

Abstract: Transformer models have become increasingly popular in financial applications, yet their potential risk making and biases remain under-explored. The purpose of this work is to audit the reliance of the model on volatile data for decision-making, and quantify how the frequency of price movements affects the model's prediction confidence. We employ a transformer model for prediction, and introduce a metric based on Partial Information Decomposition (PID) to measure the influence of each asset on the model's decision making. Our analysis reveals two key observations: first, the model disregards data volatility entirely, and second, it is biased toward data with lower-frequency price movements.

Submission Number: 48