Social Services Simulator

Project Team

Bart Gajderowicz
Michael Grüninger 
Mark S. Fox



BRAMA: Bounded Rational Methodology Architecture
We've completed the BRAMA client simulation environment, based upon theoretical foundations and methods used by social service providers.

Evaluating the Housing First Intervention Policy
Our objective is to identify key client characteristics by analyzing live data from clients participating in Toronto's Housing First program. This analysis selects key client characteristics implemented in BRAMA agents in order to emulate homeless clients.

Mapping Vocabularies
Mapping psychology vocabulary to AI planning vocabulary

The Problem
New processes for the delivery of a city's social services are often introduced based on theories or policies that have not been tested.  Hence, the recipients of these services end up being the "guinea pigs."  There are two problems with this approach. First, we do not really understand how the new policies will affect the recipients of these services.  Second, we face difficulty in understanding the results due to variations in the implementation, recipients, and service providers.
The goal of this project is to enable the evaluation of new social service policies and processes before they are subject to live trials.  In particular, to determine how changes in policies affect different types of clients of social services by means of simulation.
1. Develop a high fidelity reasoning architecture that is capable of representing a client’s decision making process.
2. Develop a process which evaluates the created models. In order to verify the high fidelity of client models, they will be incorporated into a simulation environment.
3. Use the high fidelity client models and simulation environment to evaluate the impact of various policy changes.
4. Identify how service provisioning can be improved by analyzing the client’s unique set of characteristics and constraints, as represented by the high fidelity models, and how these unique models progress through the simulations.
To develop a high fidelity model of a client's behaviour, a combination of theories is being utilized. This includes the basic Belief, Decision Intention model, extended with the TOVE ontology for representing goals, actions, states and processes, and constrained by bounded rationality and the basic needs defined by Maslow's hierarchy.
The simulation environment is based on multi-agent simulators.
Expected Outcomes
The result of this research will be a social services policy evaluation system that can be used to test new policies on simulated clients. The evaluation system will be of sufficient fidelity to predict outcomes.
Project Participants
  • Centre for Social Services Engineering
  • Factor-Inwentash School of Social Work
  • Department of Psychiatry