Son's research focuses have always been centered around logic-based artificial intelligence and its use in practical applications that can take advantage from research results in knowledge representation and reasoning, reasoning about actions and changes, automated planning, multi-agent systems, logic programming, and commonsense reasoning. He is most interested in developing logic based methodologies and systems that can answer the questions “Why is a dynamic system in its current state?” (diagnosis), “How can a system reach a desirable state?” (planning), “What if a sequence of actions is executed in the current state?” (prediction), “Which plan is best for a particular purpose?” (preferences), and "Why should one do this action instead of that action?". His current research focuses are:
- development of a high-level declarative language for representing and reasoning about multi-agent systems – the language should allow for an agent to reason about effects of actions on its knowledge and beliefs as well as knowledge and beliefs of other agents; it should also consider actions typically presented only in multi-agent domains such as lying or misleading actions;
- use of declarative technologies (answer set programming) in smart-grid/smart-home domain, nuclear reactor design, supply chain management, and multi-agent path finding problems (e.g., scheduling, planning, and diagnosis) by developing (a) automated planning/diagnostic systems for problems with multiple agents that also take into consideration agents’ knowledge, beliefs, and preferences and (b) platforms for solving multi-agent planning and diagnosis problems; and
- generation of explanations and their natural language representation to human users of the aforementioned automated planning/diagnostic systems.
His research has been supported by NSF, DoE, NIST, and NASA.