The funding for your project from Russian Science Foundation (RSF) has recently been extended, but when did you receive your first grant? And what were the extension conditions?  

We secured the grant in 2017 in the first competition of laboratories held within the RSF Presidential program. Only two projects were named winners in the category Mathematics, Informatics, and Systems Science: a project on continuum dynamics modeling in oil and gas production and our initiative on modeling processes in behavioral economics. 

The funding is not extended automatically, there is a competition. A group of specialized experts determines winners. They take into account independent reviews, the quality of reports on previous stages of the project, as well as publications on the project in highly ranked journals. The only formal condition is having an industrial partner who is ready to co-fund your research. For us, this trusty partner in both parts of our project is Bank St. Petersburg. 


Can you tell us more about the project? What is its main aim?  

In 2017, at the start of the project, we aimed at eliminating the technological divide between two financial modeling paradigms. On the one hand, there is traditional research in the field of financial mathematics with a solid formal toolset. Such works are closer to mathematics than to finance. However, this idealistic approach of this field can’t always satisfy the demands of real-life consumers.

On the other hand, there is a practical side to data science which has another approach – there, models are exclusively data-driven and thus they lose the “physics” of the finance world. In our project, we tried to bring these two approaches together, keeping in mind hybrid financial models that use machine learning to describe the processes connected to the main source of uncertainty – human behavior. And I think we succeeded. 


We developed a family of methods for multiscale modeling of various processes in behavioral economics – connected to banks and retail – and brought them to a cloud-based digital platform. It helps solve different tasks in the fields of financial scoring, optimizing a bank’s business processes, acquiring policies, forming loyalty programs, and designing new financial products. 


What studies did you conduct last year? 

In 2020, we kept working in our main field of hybrid models for financial processes and behavioral economics in general, but we were focusing more on situations with unstable and transitive processes. For instance, various critical situations, not only financial ones. The COVID-19 pandemic let us see the instability of behavioral economics. As machine learning models learn with existing data and, so to say, predict the history, they are naturally hard to use in crises – the data changes faster than it takes to train a model. That’s why now we are trying to find a way to build the models that would work in crises – that’s the current goal of our project. 

Why is it important? 

In finance, the forewarned is forearmed logic works. It is especially true for crisis situations when you don’t know what’s coming next. People are apprehensive to act and the existing motivation tools, such as customer loyalty programs, bonuses, and discounts simply stop working because they were intended for normal conditions. That’s why it’s important for any financial organization to understand the general state of behavioral economics, predict its short-term changes, and adapt the company’s business processes accordingly. That is the reason our project can be used to create decision-support systems on various levels – from recommendations to generative systems that can help design a financial product in particular conditions using AI.  

Have the results you acquired already been applied?

All of our applied solutions have already been implemented by Bank St. Petersburg, some of these solutions are used 24/7. It was one of the co-funding conditions and we don’t regret it in the slightest – implementation takes time and effort, but it also sets us new research tasks.


What are your plans for the future given the extended funding?  

We focused on five main fields that are important for work with unstable financial processes in crises. 

First of all, we build models to predict the limit states of behavioral economics and thus understand how different factors affect financial processes at a global scale. 

Second, we develop model training methods suitable for unstable data – when it changes faster than the model reaches the necessary precision level. 

Third, we work with short-term prediction methods for such processes.

Fourth, we develop generative design methods that allow us to synthesize financial products that will be efficient specifically in conditions of crisis, even if for a short time. 

Fifth, we improve tools for experimental studies in this field. As each crisis is unique and the 2020 data from the COVID-19 are clearly not enough to make general conclusions, we are creating an experimental stand based on the digital avatars (virtual personal assistants) technology, where we can hold experiments evaluating the factors of financial behavior of actual people. The experiments, however, do not involve any financial risks, as they are essentially a multiplayer computer game. This seems to be the only way to get consistent data on the diversity of crisis situations, and we find it crucial to validate all solutions suggested in our project. 

What is the ultimate goal of your project? 

The goal of any RSF project is simple – to create and share new knowledge that identifies solutions of a specific relevant task, a task that is either important for the development of the field of the project or for obtaining new results in other fields.

However, in this case we also pursue our own goal – we want to create a stable scientific school for mathematical methods in behavioral economics and finance at ITMO. We can say that over these four years the school has been established. Now, however, it needs an impulse that will make it competitive for at least a decade. We will provide this impulse by continuing our work on the RSF project in 2021-2023.