Fair Models for Impartial Policies: Controlling Algorithmic Bias in Transport Behavioural Modelling

Abstract

The increasing use of new data sources and machine learning models in transport modelling raises concerns with regards to potentially unfair model-based decisions that rely on gender, age, ethnicity, nationality, income, education or other socio-economic and demographic data. We demonstrate the impact of such algorithmic bias and explore the best practices to address it using three different representative supervised learning models of varying levels of complexity. We also analyse how the different kinds of data (survey data vs. big data) could be associated with different levels of bias. The methodology we propose detects the model’s bias and implements measures to mitigate it. Specifically, three bias mitigation algorithms are implemented, one at each stage of the model development pipeline—before the classifier is trained (pre-processing), when training the classifier (in-processing) and after the classification (post-processing). As these debiasing techniques have an inevitable impact on the accuracy of predicting the behaviour of individuals, the comparison of different types of models and algorithms allows us to determine which techniques provide the best balance between bias mitigation and accuracy loss for each case. This approach improves model transparency and provides an objective assessment of model fairness. The results reveal that mode choice models are indeed affected by algorithmic bias, and it is proven that the implementation of off-the-shelf mitigation techniques allows us to achieve fairer classification models.

Travel Demand Forecasting: A Fair AI Approach

“…The standard method of estimating travel demand is the four-step model, including trip generation, trip distribution, mode split and traffic assignments [51]. Accordingly, recent studies have started to examine and address the fairness concerns of travel demand forecasting problems spanning across these steps [12], [19], [21], [42], [52], [53]. Specifically, several studies focused on resolving unfairness issues for trip generation forecasting [12], [14], [21], [54].…”

“…Specifically, the authors predicted the Origin-Destination (OD) travel demand by using Multi-Objective Reinforcement Learning (MORL), where the objectives are optimizing transportation network's efficiency and mitigating the demand disparities among different population groups. Certain studies also explored the unfairness issue in travel mode split problems [13], [19]. For example, [13] studied the prediction disparities among population groups by using both a binary logistic regression and a three-layer deep neural network (DNN).…”

“…The unfair travel demand predictions may also cause AI-based traffic management systems to prioritize traffic flows in ways that benefit advantaged areas over disadvantaged ones, which may further worsen traffic congestion in disadvantaged communities [18]. Over time, these ill-informed transportation policies or decision-making caused by unfair travel demand predictions could lead to further unintended consequences for transportation equity [13], [19]. Accordingly, it is crucial to incorporate fairness into AI-based travel demand forecasting models [20].…”