We had the pleasure to exchange with Sébastien Touzé, digital transformation pilot at the Maritime Affairs Department: an engineer who breaks the stereotype of the introverted and laconic computer scientist.

So we took the opportunity to discuss many interesting topics: teleworking, meeting facilitation, free software, open data, business intelligence…

Chances are you will find his interview as inspiring as we do.

Tell us about yourself and your background

I’ve been working for a year at the Maritime Affairs Directorate as a digital transformation pilot. I’ve been working for a year at the Maritime Affairs Directorate as a digital transformation pilot. This position follows my recruitment as a General Interest Contractor (GIC), still within Maritime Affairs.

The EIG program is managed by the interdepartmental Digital Directorate, and recruits technical profiles within an administration for 1 year, to solve business challenges using digital tools. In particular, the aim is to optimize the use of the data available to the administration, in order to make the most of it.

The aim of my first project was to predict safety risks on ships. Indeed, the vessels of the French fleet are regularly visited in an exhaustive manner to assess their condition. Our objective was to equip the transition from systematic control of all vessels to targeted control only on those vessels that are most at risk. And this was achieved thanks to data science, i.e. algorithms.

We have used and optimized them to ensure that these algorithms best predict the need for a visit, while remaining understandable to the agents carrying out the visits. This is called “algorithm accountability”. It is about making the algorithm transparent and understandable to everyone. Following this first one-year project, I was therefore recruited to extend this type of project to other areas of Maritime Affairs.

How were the meetings within the ISG in 2019, and how are they going at the Directorate of Maritime Affairs today?

Within the GIT (General Interest Contractor) program, we did a lot of visits with the safety inspectors, the end users, to understand how they worked. Then we talked a lot with them to get their feedback on the first mock-ups.

It wasn’t always the case for geographical reasons, because the inspectors are on the French sea facades, and we are in Paris. Exchanges took place mainly by telephone. This was sometimes a limit, because sometimes we could not show an object and have an opinion on it. We were not yet sufficiently equipped at the time.

Today, my situation is different: my objective is to lead the discussion on what digital and data science could bring to the business sub-directorates of Maritime Affairs. Rather, it’s more like workshops for reflection and ideation, on the needs and difficulties that teams may have on a daily basis. For this, I use a videoconferencing tool from the State (based on an open source solution), and Tixeo, another French videoconferencing tool.

Is interaction between project participants encouraged? If yes, how?

Interaction among participants is strongly encouraged. It is essential to have a visual support, especially since these exchanges take place today in visio. To do this, I use Klaxoon to animate the meeting and get people involved, depending on how comfortable they are with digital technology (if they are more or less at ease with it).

The ideal would be to create an offline community around our exchanges to collect feedback, and to have an information sharing platform that can be animated, even when the meeting ends. It would also allow me, as a facilitator of these exchanges, to know if what we are building is really going to have value for users.

If the health situation permits, would these meetings take place in-person?

The benefit of COVID is that it has forced teleworking and accelerated the development of telepresence tools.

If it weren’t for COVID, our exchanges would be a mixture of face-to-face and distance, because it’s important that these reflections on digital tools involve everyone, including those who are not in Paris.

Open source software and data security: how to limit the risks of cybercrime attacks?

Free software is not less secure than proprietary software, on the contrary!

If the source code is not open, explained and understandable, how can I be sure that the algorithm makes the decisions it is supposed to make? Open source software makes it possible to analyze the code, to evaluate its durability, and to perform a quality and security audit: to know if there is a way to penetrate (whether in the stored data, or in the exchanges, in the case of a videoconferencing tool).

The opening of the code makes it possible to multiply the tests and prevent cyber attacks: anyone can analyze and report the risk of possible technical flaws. The essential condition is that these remarks are taken into account by the person who keeps the software alive, to make it evolve through corrections. It is also important to bear in mind that many closed source software are strongly built from open source software.

We often hear about the importance of open data for public administration. Could you give us an example of this?

Access to public data of the administration is a right expressed in Article 15 of the Declaration of Human Rights.

This is one of the reasons why the GIS program is strongly imbued with the notion of open data, the sharing of data. Let us take the example of court decisions: after having anonymized certain information, sharing makes it possible to have a repository accessible to everyone.

The same goes for information on the French vessel fleet, which can be used for economic statistics or other purposes.

How has BI (business intelligence) evolved in recent years, and what are, in your opinion, the trends for BI in 2021?

Today, at the Maritime Affairs Directorate, we have a lot of indicators that are monitored with BI tools.

The difficulty is that these tools are often very static: you define an indicator, and you’re not supposed to touch it too often. However, public policies evolve much faster than the ability to make indicators evolve in BI tools.

The aim is to move to more agile tools that allow users to explore, understand and appropriate data. And beyond that, it’s a question of relying on data science with forecasting tools.

In this way, we can predict the evolution of the indicator by exploiting a mass of information that goes beyond our administration’s own data, such as global economic data, for example. To sum up: making sure that data is no longer “just a computer scientist’s trick!

Interview conducted by Laura Mattiucci