Unlocking the Black Box: Explainable AI in Resource Scheduling

Why do you need Explainable AI?

At Solvice, we've long recognized the role of transparency in AI Optimization. This has led us to discuss and evaluate different forms of Explainable AI for optimization with our customers. We came to three important realizations:

• Transparency: It is imperative to gain insight into how solvers make decisions. This lays the groundwork for a deeper understanding and fine-tuning of the solutions our customers build.
• Debugging: By clarifying the decision-making process, developers can identify and rectify parameterization issues more effectively.
• Trust: Explainable AI demystifies the solver's logic, fostering a stronger sense of reliability among users. This trust in our technology drives adoptions and avoids churn, ensuring you can rely on Solvice's Explainable AI feature for your AI scheduling needs.

Historically, our approach to shedding light on decision-making was to provide a list of violated constraints in a solution. Yet questions persisted, calling for a more nuanced explanation of solver choices, from shift assignments to route planning.

How It Works: Explaining the Unexplainable

Prompted by customer inquiries that sought clarity on specific solver decisions, we started creating a feature that transcends traditional explanations. This gave birth to the first version of our Explainable AI feature, a tool that provides detailed insights into our solvers' decision-making process. This led to the “explain endpoint,” now in beta and available for our FILL API and VRP API solvers.

Our new feature is based on a hyperlocal discovery phase, which is a detailed examination of possible alternatives after identifying the optimal solution. This phase uncovers all feasible alternative assignments for every decision rendered by the solver. So if someone wonders why a specific resource has been assigned to a job, Explainable AI will tell the user how good the schedule would be if the user were to choose one of those alternatives. Given the computational intensity of this task, it's triggered only upon user request for an explanation.

In summary, this exhaustive search yields a score evaluation, which is a measure of the quality of each alternative assignment for potentially countless alternatives. This gives users the necessary insights to understand why certain constraints were violated and how minor adjustments could enhance the solution.

In Action: The Shift Scheduling Example

Consider a scenario in shift scheduling where our solver not only assigns shifts optimally but also clarifies why each shift was allocated to a particular employee. By exploring alternative shift assignments and evaluating their impact on the overall solution, users are armed with the knowledge to make informed adjustments.

The Future with Explainable AI

As we roll out this feature, we invite developers and product managers to experience firsthand the transformative power of Explainable AI. For those constructing their own scheduling software, Solvice's latest innovation offers not just a tool but a collaborative partnership in pursuit of greater efficiency, transparency, and trust.

To explore the potential of Explainable AI in your scheduling solutions, we encourage you to dive into our API reference, which provides detailed information on integrating our Explainable AI feature into your software, and our comprehensive guides, which offer step-by-step instructions on how to use the feature. The future of resource scheduling is not only intelligent but understandable. Welcome to a new era with Solvice's Explainable AI.

Get started with Explainable AI for Scheduling

‍