AI Requirements for Manufacturing in the XMANAI Project
The XMANAI project is working to provide the tools to navigate the Artificial Intelligence (AI)’s “transparency paradox”, designing, developing and deploying a novel Explainable AI Platform powered by explainable AI models that inspire trust, augment human cognition and solve concrete manufacturing problems with value-based explanations. The platform needs to be fully aligned with the manufacturing needs and idiosyncrasy, acting as a single reference point of access both for AI and for manufacturing value chain stakeholders, providing a collaborative environment for Data Scientists and Manufacturing Experts to create machine learning flows for modeling manufacturing data.
To ensure that such high-level objective meets the requirements of the community, the team is implementing an agile methodology for requirements elicitation, promoting interactive sessions with stakeholders, that have been very enlightening and that can serve as a case study for other projects. Let’s have deeper look into the process we are following.
Requirements in an Agile process may come in many shapes and forms, but the most common is through User Stories.
What is a User Story?
A User Story is a requirement expressed from the perspective of an end-user goal. User Stories represent the needs of the customer in a simple written narrative instead of a boring comprehensive document that most of the time fail to provide context.
User Story descriptions typically follow a simple template as a Card:
- As a < role>,
- I want < goal >
- so that < benefit>
Why to use User Stories?
The main principle behind User Stories is that the product could be fully represented through the needs of its users. In fact, User stories are short and simple descriptions of a feature told from the perspective of the person who desires the new capability. Their focus is on why and how the user interacts with the system, which provides a great deal of context to developers working on the system specification. A User Story is essentially a high-level definition of what the system should be capable of doing from the perspective of the people that will actually be using it, fitting perfectly in SCRUM agile frameworks.
User Story Mapping
User Story Mapping is an Agile approach to arrange the User Stories in two dimensions:
- Vertical: demonstrating the priority of the User Stories for different system releases and MVPs
- Horizontal: representing the user journey (steps a user takes to perform actions in the system) and grouping User Stories into higher levels of functionality/activity.
The User Story Mapping process helps to present user stories into a useful model for understanding the overall functionality of a system.
How user stories are used to extract XMANAI requirements
Requirements of the XMANAI platform are extracted in the form of user stories, collaboratively by technical partners and business stakeholders through interactive sessions, which in the current days with the Covid-19 pandemic, required a bit more imagination and organization.
XMANAI has been using Miro(www.miro.com), which is an online collaborative whiteboard platform, configured with separate User Story boards for each of the stakeholder roles in XMANAI, namely Data Scientist, Data Engineer and Business Expert. The interaction between the partners to fill-in the User Story boards stimulates creativity and provides the opportunity to examine the needs not only from a technical point of view but also from a practical point of view, defining the behavior of the XMANAI platform towards the initial goal.
Several user stories are extracted at the end the brainstorming, and the user journey specification may follow. The journey of each user is broken into the phase of activities and organized as the sequence of all the possible events that a user goes through. The extracted user stories are then mapped and assigned to a related phase of activity as in the figure bellow.
After several interactive sessions and further discussion, a first version of the technical requirements is now available and ready for the development team to exploit. The high-level XMANAI AI requirements for manufacturing are summarized next.
As an XMANAI user (data scientist and manufacturing business experts), I want to …
- have a collaborative environment where data scientists and business users will be able to work together to define AI and machine learning algorithms.
- have tools for Data Preparation and Manipulation (cleaning, completion, enrichment, linking and transformation) to make input data appropriate for be used by the AI models and machine learning pipelines.
- execute AI Models, Visualise and extract the results over both on-premise and cloud-based infrastructure.
- have prescriptive analytic methods, to allow the extraction of patterns and the execution of “what-if” simulation scenarios based on the existing data that is available.
- have explainable AI models and Graph machine learning Algorithms to be integrated in data analysis in order to gain explainability in AI solutions.
- to be able to validate my AI models through cross-validation methods and expert scoring for the accuracy, performance, and value of such models, by combining the knowledge of data scientists and business experts.
- have a secure and trusted environment for my data, models and results that are employed on cloud and on-premise.
Such requirements go in the direction of the expected goal for the XMANAI platform, providing a secure and collaborative environment for manufacturing experts and data scientists to take the advantage of Artificial intelligence for the progress and enhancement of their businesses. XMANAI will help the manufacturing value chain to shift towards the amplifying AI era by coupling (hybrid and graph) AI “glass box” models that are explainable to a “human-in-the-loop” and produce value-based explanations, with complex AI assets (data and models). With XMANAI, Business Experts will be able to extract manufacturing-relevant data from the manufacturer’s backend systems via appropriately defined APIs, and through different data mapping and data semantic annotation techniques, they can attach significant added value to their data and increase data quality, all while safety is ensured via the XMANAI attribute-based access control policy framework.
