It’s undeniable that Artificial Intelligence (AI) has become part of everyone’s life. It is used by companies to exploit the information they collect to improve the products or services they offer and, wanted or unwanted, it is present in almost every device around us.
Not surprisingly, AI is also starting to impact all aspects of the system and software development lifecycle, from their upfront specification to their design, testing, deployment and maintenance, with the main goal of helping engineers produce systems and software faster and with better quality while being able to handle the ever increasing complexity of software-intensive systems.
In the last years, the MDE community has taken the first steps working on the integration of AI techniques into modeling techniques, technologies, and processes. For example, under the umbrella of the Modelia project — a joint collaboration between CEA LIST and UOC — we have contributed to the state of the art in different directions such as the automatic inference of model transformations  and the development of a low-code framework for the creation of chatbots . Another area where AI and MDE have been combined is in the application of search-based approaches, e.g., using genetic algorithms, in an MDE context: Several approaches now exist allowing to search for optimal model transformations  and optimal models . The field is big enough for a first survey to have been produced , a TTC case , and, more recently, we have produced the first comparative study of search-based approaches in MDE .
We truly believe that Model-driven Engineering (MDE) and Artificial Intelligence (AI) can clearly benefit from cross-pollination and collaboration. There are at least two ways in which such integration—which we call MDE Intelligence—can manifest:
- Artificial Intelligence for MDE. MDE can benefit from integrating AI concepts and ideas to increase its power: flexibility, user experience, quality, etc. For example, using model transformations through search-based approaches, or by increasing the ability to abstract from partially formed, manual sketches into fully-shaped and formally specified meta-models and editors.
- MDE for Artificial Intelligence. AI is software, and as such, it can benefit from integrating concepts and ideas from MDE that have been proven to improve software development. For example, using domain-specific languages allows domain experts to directly express and manipulate their problems while providing an auditable conversion pipeline. Together this can improve trust in and safety of AI technologies. Similarly, MDE technologies can contribute to the goal of explainable AI.
With the goal to provide a forum to discuss, study and explore the opportunities and challenges raised by the integration of AI and MDE, and broaden the community in the field, this year, we are organizing the 2nd edition of the MDE Intelligence workshop, co-located with MODELS 2020 in Montreal, Canada. We’re inviting you all to join us for this workshop and submit your papers and ideas.
After reading, brainstorming and discussing, we have come up with an initial set of topics that we believe are worth exploring:
- AI for MDE
- Application of (meta-heuristic) search to modelling problems;
- Machine learning of models, meta-models, concrete syntax, model transformations, etc.;
- AI planning applied to modelling, meta-modelling, and model management;
- Modeling assistants such as bots, chatbots and virtual assistants/recommenders supporting diverse modeling tasks;
- Model inferencers and automatic model generators from datasets;
- Self-adapting code generators;
- AI-based user interface adaptation for modeling tools;
- AI with human-in-the-loop for modeling;
- Semantic reasoning platforms over domain-specific models;
- Semantic integration of design-time models with runtime data;
- General-knowledge or domain-specific ontologies;
- Probabilistic models;
- Use of AI techniques in data, process and model mining and categorisation;
- Natural language processing applied to modelling;
- Perception and modeling.
- MDE for AI
- Domain-specific modelling approaches for AI planning, machine learning, agent-based modelling, etc.;
- Model-driven processes for AI system development;
- MDE techniques for explainable AI;
- Using models for knowledge representation;
- Code-generation for AI libraries and platforms;
- Model-based testing of AI components.
- Tools for combining AI and MDE;
- Case studies in MDE Intelligence;
- Challenge problems to be addressed by combining AI and MDE techniques.
- AI for MDE
If you’re interested in contributing or participating either actively or passively in the workshop, check the workshop website, leave a message below or contact us directly.
We’re looking forward to hearing from you!
 Loli Burgueño, Jordi Cabot, Sébastien Gérard: An LSTM-Based Neural Network Architecture for Model Transformations. MoDELS 2019: 294-299
 Gwendal Daniel, Jordi Cabot, Laurent Deruelle, Mustapha Derras: Xatkit: A Multimodal Low-Code Chatbot Development Framework. IEEE Access 8: 15332-15346 (2020)
 MOMoT: http://martin-fleck.github.io/momot/
 MDEOptimiser: https://mde-optimiser.github.io/
 I. Boussaidet al., “A survey on search-based model-driven engineering”, Automated Software Engineering, vol. 24, pp. 233–294, Jun 2017.
 Stefan John, Alexandru Burdusel, Robert Bill, Daniel Strüber, Gabriele Taentzer, Steffen Zschaler, and Manuel Wimmer: Searching for Optimal Models: Comparing Two Encoding Approaches. 12th International Conference on Model Transformations (ICMT’19), 2019.
Lola Burgueño is a postdoctoral researcher in the SOM Research Lab at the Internet Interdisciplinary Institute (IN3) of the Open University of Catalonia (UOC), in Barcelona (Spain) and CEA List in Paris (France). Her research interests focus on Software Engineering (SE) and Model-Driven Engineering (MDE). She has and is contributing to the fields of capturing and operating with uncertainty in software models for its use in the Industry 4.0, AI-enhanced software systems, testing models and transformations, the distribution of very large models and the parallelization of their manipulation to boost performance, among others.