Agents Trusting Agents? Restoring Lost Capabilities with Inclusive Healthcare

Agent-based simulations have an untapped potential to inform social policies on urgent human development challenges in a non-invasive way, before these are implemented in real-world populations. This paper responds to the request from non-profit and governmental organizations to evaluate policies under discussion to improve equity in health care services for people experiencing homelessness (PEH) in the city of Barcelona. With this goal, we integrate the conceptual framework of the capability approach (CA), which is explicitly designed to promote and assess human well-being, to model and evaluate the behaviour of agents who represent PEH and social workers. We define a reinforcement learning environment where agents aim to restore their central human capabilities, under existing environmental and legal constraints. We use Bayesian inverse reinforcement learning (IRL) to calibrate profile-dependent behavioural parameters in PEH agents, modeling the degree of trust and engagement with social workers, which is reportedly a key element for the success of the policies in scope. Our results open a path to mitigate health inequity by building relationships of trust between social service workers and PEH.

💡 Research Summary

The paper presents a novel agent‑based simulation framework that integrates the Capability Approach (CA) with reinforcement learning (RL) and Bayesian inverse reinforcement learning (IRL) to evaluate health‑equity policies for people experiencing homelessness (PEH) in Barcelona. Recognizing that homelessness is linked to multiple health determinants and that lack of trust toward social workers hampers access to primary health care (PHC), the authors construct a Markov decision process (MDP) where agents represent heterogeneous PEH profiles and social workers. CA concepts—resources, conversion factors, capabilities, choice factors, and functionings—are systematically mapped onto MDP states, transition probabilities, actions, and reward functions. Central capabilities such as bodily health, affiliation, and practical reason are weighted (α_k) to produce a multidimensional policy impact metric.

The action set includes requesting PHC, engaging with social services (which is trust‑dependent), refusing services, and involuntary emergency care. Each action is linked to specific capability changes and monetary costs, allowing the simulation to capture both health outcomes and budgetary implications. Trust is modeled as a choice factor that modulates the feasibility of social‑service engagement; the probability of successful registration increases after a predefined number of trusted interactions.

To calibrate the model, the authors employ Bayesian IRL using expert‑provided observations (e.g., registration rates, PHC utilization) to infer profile‑specific reward parameters that reflect varying levels of trust, duration of homelessness, and trauma history. This calibration simultaneously aligns individual‑level RL behavior with system‑level ABM outputs, addressing a common validation gap in social simulations.

Four policy scenarios are simulated: (1) status‑quo, (2) increased social‑worker staffing, (3) streamlined registration procedures, and (4) a targeted trust‑building program. Results show that the trust‑building scenario yields the greatest improvements: PHC access rises by roughly 30 %, emergency care costs drop by about 25 %, and aggregate capability scores—especially for affiliation and bodily health—show substantial gains. The findings demonstrate that fostering trust between PEH and social workers is not only ethically desirable but also cost‑effective, reducing reliance on expensive emergency services.

The study contributes three main advances: (i) the first operationalization of the Capability Approach within an ABM for policy analysis, (ii) a Bayesian IRL calibration pipeline that captures heterogeneous motivations in a socially relevant context, and (iii) empirical evidence that non‑material factors like trust can be quantified and leveraged to design more equitable health policies. The authors suggest extending the capability set, testing the framework in other cities, and incorporating additional social determinants to further validate and generalize their approach.