Understanding and modelling the impact of consumer purchasing behaviour on the global supply chains' decisions in adapting anti-slavery practices

The growth of global supply chains has caused an asymmetrical distribution of risk and profit within the chains where chain captains - often end buyers, such as retailers - dominate supplier relationships. As a result, risk is pushed down the chain to lower-tier firms, which have to
address increased uncertainties and consequently, to workers who have to shoulder the burden of required flexibility. Global supply chains driven by the principle of comparative advantage find their way to networks where lower-tier firms are able to reduce their cost structure by resorting
to forced labor practices. Power relationships and fragmentation of global supply chains call for adopting a shared responsibility approach, involving participation and leverage of stakeholders.
The UK Modern Slavery Act formally assigns consumer a critical role in eradicating modern slavery not only through actual demand but also as a key stakeholder who represents other stakeholders in the socio-political scene [5, 6]. The demand-side management of anti-slavery
practices is still an under-researched academic field. It highlights the need to investigate the contextual (e.g. behavioral biases and social influence) and psychological (e.g. neutralization and legitimation) barriers to consumer action, assess characteristics of consumer segments, identify influential factors, and develop segment-specific strategies to nudge behavior towards a slavery free world. Practical implementation of these strategies depends highly on the accuracy of demand forecasts at different levels of the supply chain network characterized by complex
hierarchies with different levels of aggregation.
The application of AI, especially machine learning methods, in tackling modern slavery offers numerous opportunities for theoretical and empirical pathways. These methods allow for the incorporation of behavioral considerations into decision-making by using a large number of
observations and attributes. They also prove promising for the development of accurate hierarchical forecasting models exploiting the benefit of multi-variate data.