My team at IBM Research has created a unique tool, called IBM Research Scenario Planning Advisor, that can use AI planning to support risk management activities in areas like security and finance. IBM Research Scenario Planning Advisor is a decision support system that allows domain experts to generate diverse alternative scenarios of the future and imagine the different possible outcomes, including unlikely but potentially impactful futures.
Planning and plan recognition
Preparing for the future is fundamental to the success of most human endeavors, from playing chess to running a multinational organization. At IBM Research AI, we build AI-based systems that use expert knowledge and AI planning to reason about observations derived from relevant news and social media and generate explanations and hypotheses about the current state of the world—and many possible alternative future states.
Planning is a long-standing area of research within AI. Planning is the task of finding a procedural course of action for a declaratively described system to reach its goals while optimizing overall performance measures. AI planning can help when (1) your problem can be described in a declarative way; (2) you have domain knowledge that should not be ignored; (3) there is a structure to a problem that makes it difficult for pure learning techniques; or (4) you want to be able to explain a particular course of action the system took. A plan recognition problem is the inverse of a planning problem: instead of a goal state, you are given a set of possible goals. The task in plan recognition is to find out which goal was being achieved and how.
Scenario planning is a widely accepted technique by which organizations develop their long-term plans. Scenario planning for risk management puts an added emphasis on identifying the extreme yet possible risks and opportunities that are not usually considered in daily operations. Scenario planning involves analyzing the relationship between forces (such as social, technical, economic, environmental, and political trends) in order to explain the current situation, in addition to providing insights about the future. This process is depicted in the picture below.
A major benefit to scenario planning is that it helps us to learn about and anticipate possible alternative futures. We use scenario planning because we cannot predict the future. We use AI planning, informed by expert domain knowledge, because some scenarios have never yet occurred and thus cannot be projected by probabilistic means. And we generate many different scenarios, exploring a variety of possible futures, because we want to be prepared for both expected and surprising futures.
IBM Research Scenario Planning Advisor
Our approach transforms risk management into a plan recognition problem and applies AI planning to generate solutions. It addresses several challenges inherent to this task. They include: (1) having inconsistent, missing, unreliable observations; (2) being able to generate not just one but many future plans; and (3) being able to capture and encode the necessary domain knowledge.
IBM Research Scenario Planning Advisor includes tooling for experts to intuitively encode their domain knowledge and uses AI planning to reason about this knowledge and the current state of the world, including news and social media, when generating scenarios. In our recent paper at the 2018 Association for the Advancement of Artificial Intelligence (AAAI) conference , we first characterize the scenario planning problem as a plan recognition problem and then use AI planning to generate many possible plans. Finding one plan is computationally challenging (it is PSPACE-complete), but our system finds a set of plans. We transform the domain knowledge into a planning task, the risk drivers into observations, and the business implications into the set of possible goals. We then use planning to compute a set of plans. We cluster these plans and present a handful of scenarios to the users.
Our system can be applied in network security, healthcare, and finance, enterprises which have at least two factors in common: (1) they have teams of analysts and domain experts who can provide the necessary domain knowledge to the system, (2) they generate many news events that can serve as observations of their current states and data points for where they are headed in the future. Our system is able to explain the past and project the future by providing a range of possible scenarios and an explanation for each scenario.
Planning for risk management
We currently have focused on applying our approach to scenario planning for risk management. IBM Research Scenario Planning Advisor is currently in deployment within IBM, supporting financial teams in their risk management activities. The system’s cognitive tools assist analysts in two ways. First, it provides situational awareness of relevant risk drivers by detecting emerging storylines. Second, it automatically generates future scenarios that allow analysts to reason about, and plan for, contingencies and opportunities in the future.
The picture below shows an example of a scenario the system produces. Each scenario we produce highlights: (1) the potential leading indicators, the set of facts that are likely to lead to a scenario; (2) the scenario and emerging risk, the combined set of consequences in that scenario; and (3) the business implications, a subset of potential effects of that scenario that the decision-makers care about.
Two years in, and the MIT-IBM Watson AI Lab is now engaging with leading companies to advance AI research. Today, the Lab announced its new Membership Program with Boston Scientific, Nexplore, Refinitiv and Samsung as the first companies to join.
IBM researchers published the first major release of the Adversarial Robustness 360 Toolbox (ART). Initially released in April 2018, ART is an open-source library for adversarial machine learning that provides researchers and developers with state-of-the-art tools to defend and verify AI models against adversarial attacks. ART addresses growing concerns about people’s trust in AI, specifically the security of AI in mission-critical applications.
It is no surprise that following the massive success of deep learning technology in solving complicated tasks, there is a growing demand for automated deep learning. Even though deep learning is a highly effective technology, there is a tremendous amount of human effort that goes into designing a deep learning algorithm.