Crossing Viewpoints: Data & Humanism

Authors:

●       Michel LUTZ Group Data Officer, Total / Research Fellow, LIMOS, [email protected]

●       Jér?me BRUNEEL, Co-Founder & CEO of Sphères (innovation & design firm), [email protected]

To quote this article :

Lutz, M., & Bruneel, J. (2019). Croisement de points de vue : données & humanisme. Management & Data Science, 3 (1).

The following opinion piece was written by two experts. It springs from the idea of comparing contrasting points of view between M. Lutz, Group Data Officer at TOTAL, and J. Bruneel, an entrepreneur in SPHERES, a design & innovation firm that is pioneering design-data approaches in France. Together, they advocate for a humanistic approach to the development of algorithms.

If working on data science projects has taught us one thing, it’s that all too often we have found ourselves dazzled — or worse, blinded — by those projects.

Dazzled indeed, as those projects allow us to build models that go far beyond our ability to interpret, with their strong nonlinearities and high dimensions. Faced with such potential, we are often left, at the end of the projects, with two opposing choices:

1) oversimplifying the models in order to make them understandable to the "professional sectors", which ultimately diminishes their value;

2) coming to terms with using a model that we do not fully understand, thus creating "zombie-like" decision systems, at odds with the human knowledge system.

Neither outcome is satisfactory.

Blinded, then: For believing too much in the power of data, we tend to forget the rest. However efficient those projects’ performances are, reality will always end up being more complex than what was measured. Deciding based solely on data is a terrible amputation of our decision-making power and betrays a lack of confidence in our ability to understand the world.

What we need is to bring people closer to data and models, so that we can better understand the world and make the right decisions. We believe that talking about humanism will enrich the conversation on the "ethics of artificial intelligence" we keep hearing about.

This humanist dimension can take various shapes and forms, whether it is through collaboration, the meaning we assign to data-based projects, the way we understand systems, or our reliance on the intuition / verification / improvement process ...

For some years now, data has been approached by companies as an opportunity to:

●       better understand the behavior of human beings (customers, collaborators, partners ...),

●       steer the company with greater precision — including its assets, operations, industrial tools …

●       build new business models, often adjacent to historical activities by enhancing companies’ ability to capture, analyze and return that asset

But what we have observed in recent years too is a technological arms race, as well as a "use cases" race to rank among the pioneers of data. Technology has therefore tried to join with professional expertise — but unfortunately, we have often left behind what binds them within a company: human beings.

Whatever the scenario, it always begins with the need, expressed by a human being, who must then apply various skills and practices to achieve a goal by articulating, once again, human networks.

But many obstacles lie ahead.

Human beings must be able to express themselves, regardless of their expertise, in an atmosphere of collaboration — by combating assumptions and demystifying certain beliefs or fears surrounding data, caused by a poor understanding of its use, its possibilities, its sources or its technological challenges.

Such a "data-humanism" thus revolves around a common language and a common multidisciplinary approach, established by the data technicians (data scientists, data engineers ...), on-the-ground operators of the company’s trades, designers, strategists, jurists ...

It seems then restrictive to only speak about data, as the stakes now also include the transformation of exchanges between the different actors of a project and the emergence of new tools and approaches.

This is where matters of meaning and form become crucial:

●       How does what we are building serve humankind, in the broad sense?

●       In what way will the humans who will use these new services, these piloting tools, these visualizations, find a concrete and practical use in their environment?

●       How will models and concepts be translated into more enlightened and concrete operating actions?

●       How does all of this augment workers and allow them to gain an even finer understanding of the complex world in which they’re evolving?

In short, our use of data must place humans at the center of the issue and help them in their daily tasks and activities. And above all, it must complete (without replacing) our intuition — that inherently human capacity, separate from all predictable logic and often at the center of entrepreneurial activities.

In that respect, the notion of "augmented human" takes on a different meaning than the one we are used to, which merely focuses on automating some of our repetitive or tedious activities.

Much more than that, a humanistic approach to data makes it possible to rethink the relation between man and his daily life; so that he may become even more human, more free to rely on his natural intuition, his sensitivity, his inner heart, his spirituality.

As in music, wherein mastering technique can have a liberating effect on creation, data can be perceived as the liberating technique of intuition.

In The World as Will and Representation, Arthur Schopenhauer illustrates this point of view: perfectly: "The starting point of all beautiful work, of all great and deep thought, is a perfectly objective intuition."

Just like the digital revolution, data therefore effectuates extremely dense transformations within companies. Artificial intelligence then presents a new challenge to rethink our ways of doing, our ways of working.

Based on this, and considering how data models evolve, as they are trained and made stronger through trial-and-error processes, we can naturally draw a "methodological" analogy with modern entrepreneurial or product development behaviors.

Our point of view stems from the need to accept, with humility, that it is by acting on data first as social, collaborative and intuitive human beings that we will contribute to the success of an algorithm, a new service or a transformation of practices.