Human-centered AI for Collaborative Policy Making

Why Interventions Fail

Policies, programs and other interventions fail to achieve results for two reasons:

They are built on guesswork, instead of the best available evidence.

People are not given the opportunity to have a meaningful say in the design of policies that affect them, and they are reluctant to accept others' imposed perspectives.

An approach to Understanding Any Intervention System

Virtually all intervention systems are governed by complex cause and effect relationships. Detailed knowledge of these relationships is a critical prerequisite for understanding and predicting how systems will behave. PolicySpark’s platform tames complexity and provides the critical strategic information required to design and implement effective and efficient interventions.

Our Software

Allows every voice to be heard

Analyzes the evidence and data to identify which policy levers produce the desired outcomes

Identifies where people agree

Advantages for Decision-makers

Incorporate the latest evidence, data science and best practices

Demonstrate that stakeholders have been listened to and are a meaningful part of decision-making

Identify which policy levers are most effective in producing desired outcomes

Implement effective policies that minimize disruption and cost

Measure desired outcomes

Implement a cycle of continuous improvement

Why Use Evidence?

Policies and Programs Informed by Evidence:

Are more likely to achieve desired outcomes

Are more likely to achieve desired outcomes

Build trust, buy-in and communication with stakeholders and benefactors

Increase institutional credibility

Satisfy the increasing demand for evidence-informed decision-making

Increase accountability

Three Steps to Better Results

Step 1:

Gather Evidence

Stakeholder Knowledge

Experiential evidence derived from deliberative stakeholder narrative

Expert Knowledge

Evidence derived from subject matter experts


Evidence derived from existing research literature

Big/Open Data

Evidence derived from available big and/or open data sets

Step 2:

Identify Key Outcome Pathways

Sources of evidence are transformed into models that reflect the many causal factors that play a part in driving outcomes. Data are combined and analyzed using PolicySpark’s algorithmic platform resulting in a model that reveals the factors that exert the highest levels of influence on desired outcomes. With these “leverage points” in hand, it becomes possible to describe detailed results pathways, providing the basis for measurable and testable theories of change.

Step 3:

Manage Interventions Strategically

By systematically gathering and processing the available evidence, decision-makers and managers are in a better position to make more informed choices. Our products can be used to:

  • Design new interventions
  • Review / re-design existing interventions
  • Identify the key factors that drive results
  • Conduct root cause analysis
  • Develop evidence-based theories of change
  • Develop evidence-based implementation plans
  • Design meaningful performance measurement systems
  • Identify meaningful performance indicators
  • Use predictive modeling to visualize what-if scenarios
  • Take an experimental approach to find out what works
  • Engage in informed, intelligent resource allocation
  • Transform evaluation activities
  • Communicate with stakeholders
  • Demonstrate accountability

Executive team


Chris Callaghan

A native Montréaler, Chris studied math, chemistry and physics at Dawson College in the mid-80’s, and went on to earn earned a B.Sc. from McGill in natural history and wildlife science, and an M.Sc. from the University of Ottawa, with a focus on how human activities affect natural ecosystems. Chris then entered the Canadian federal public service, where he took on various positions related to environmental toxicology, ecosystem management, sustainable development and environmental auditing... More Info

It was during his time with the Auditor General of Canada that Chris became interested in the policies and frameworks associated with measuring organizational performance and accountability. In 2006, Chris opened his own consulting firm specializing in program evaluation. As he gained insight into some of the challenges associated with program evaluation theory and practice, he observed what he sees as serious gaps in how interventions are designed and measured. Chris is now concerned with developing better means to understand how interventions work, how to improve design processes, and how to support evidence-informed decision-making. His vision is a world in which decision-making is systematically supported with evidence, including research, expert knowledge, stakeholder narrative, and open/big data. To realize this vision, he is centering his focus on the use of cutting-edge concepts and technologies, including those associated with data science and machine learning, to gather, analyze and test the evidence needed to bring policy-making into the modern, data-driven era.


Matthew Spencer

Matt is PolicySpark’s Principal Scientist. Matt is an applied scientist with expertise in enterprise-scale machine learning, software development, data science, and computational biology. He is the engine behind PolicySpark’s innovative technological solutions. He obtained a BASc. in computer engineering from the University of Ottawa and went on to earn a Ph.D. in cybernetics from the University of Reading in the United Kingdom. The focus of his thesis was complex network models of brain connectivity... More Info

Matt is an effective international and inter-disciplinary collaborator. He has over 10 years of experience working with teams around the world and communicating complex ideas to a wide range of audiences. He has 5 years of experience leading machine learning projects at Amazon. His technical expertise includes machine learning research and development employing a wide variety of technologies, including TensorFlow, Sklearn, Spark, Python and AWS. Other strengths include natural language processing, biosignal/time-series analysis, complex network analysis (Graph Theory), predictive stochastic models, reinforcement learning, programming (Python, Java, C#, C/C++, MATLAB), and enterprise software architecture.

Contact Us

Send us a Message