Artificial intelligence as a tool for digital inclusion

In 2018, entrepreneur and former media company director, Olivier Mégean, founded, a strategy consulting firm specialised in artificial intelligence. Contributor to the white paper “Contre l’illectronisme” [1] (Against digital illiteracy) published in 2019, he defends an inclusive vision of artificial intelligence and of machine learning.

Man-machine interactions have greatly improved in quality.

How is digital exclusion manifested today in France?

Today, the phenomenon is well-known and the studies carried out on this subject describe converging tendencies. Digital exclusion concerns slightly more than 15 % of the population. It is due to several factors. First and foremost, it is territorial: 50 % of non-internet users live in municipalities of less than 20,000 inhabitants. In addition to this are two determining factors of inequality: the levels of education and of income. And, as expected, there is a generational dimension: one in two over-75-year-olds don’t have access to the internet at their home, whereas only 2 % of 15-29-year-olds don’t.

But using the internet does not suffice to master the basics of digital. Thus, 38 % of users have trouble with some of the most “basic” usages. The term “digital illiteracy”, which today has consensus, may be disturbing, but in this respect it is very meaningful. And, as with any phenomenon of exclusion, it is not acceptable and must be tackled.

Could innovations in the area of artificial intelligence not only reduce digital exclusion but also play a part in social inclusion in a broader sense?

More so than we imagine. I’m thinking in particular about technologies relating to the understanding and generation of natural language. Voice interaction, notably, appears as a particularly pertinent solution for addressing difficulties reading and writing on digital devices (computers, tablets, or smartphones). Today, the machine is capable of interpreting the intention underlying a request or a spoken remark. The ability of administrations and of large companies to develop voice interaction possibilities with users and customers is most likely one of the conditions for a decrease in digital illiteracy in France.

Thanks to specific training, other machine learning technologies make it possible to detect atypical browsing behaviours and to identify people in difficulty browsing the internet. Today, these technologies are deployed on many sites for purposes of assistance and of recommendation. Yet it must still be possible to offer internet users a means of switching to a human interlocuter to help them finalise their action.

What role can smart assistants play?

Smart assistants were viewed at first as gadgets or spies. Today, they are settling massively into homes. Man-machine interactions have greatly improved in quality, promising virtual assistants a true functional utility, beyond fun and futile applications. What’s more, the price of the tools is no longer an obstacle to their deployment, including for economically fragile populations. And should the price remain an obstacle, one can suppose that public authorities are capable of imagining aid to enable everyone to be equipped and benefit from truly useful functions.

Some examples of useful functions?

Microsoft is working on developments that will make it possible to turn visual data into audio feedback and thus enable blind people to “listen” to photos. For its part, Amazon is fairly advanced in sign language: a camera is in charge of interpreting the gestures of a deaf or hearing-impaired person to enable a smart assistant to formulate a verbal (smart assistant) or written (computer) answer. Alexa can also provide many services to the elderly and break isolation: launching podcasts or radio programmes, various interactions (dialogue, games to stimulate the mind, sequencing of services throughout the day), voice control of home automation, relay to emergency services in case of a fall, etc.

French initiatives seem rather rare…

In order for artificial intelligence to act fully as a tool of inclusion, it first needs data, which is anonymous of course but also sufficiently neutral. The data coming from human behaviour contains an infinite quantity of biases that don’t enable true neutrality in terms of processing and response. In France, we are lucky to have all the talents required to succeed in this challenge. It is then necessary to be able to host this data in compliance with the security criteria of France’s ethical requirements, and in all cases in accordance with European democratic values. Today, this hosting space does not exist, or at totally prohibitive prices. But it will come, I have no doubt.

Read also on Hello Future

Artificial speech at Orange: fluency is the priority!