Robin: a virtual assistant that bonds with its user

By February 15, 2016
Assistant virtuel

In the same way that people tend to tailor what they say and how they say it according to whom they are talking to, a virtual assistant also needs to adapt to its owner.

In a previous article, L’Atelier reported on the virtual assistant (VA) market, potential VA applications and upcoming challenges. In this article we now take a look at the way a virtual assistant is able to bond with its owner, adapting to the user and responding optimally to his/her needs. As many people feel they are diving into the unknown with VAs, the assistant’s capacity to adapt is a key factor that may well decide whether a person will adopt the technology or not.

Robin Labs, a young Palo Alto, California-based company which specialises in developing software that can hold a conversation  with human beings, is mainly known for its flagship product, Robin, a virtual assistant designed to be ‘your personal eyes-free assistant on the road’. Robin lets you send messages, obtain traffic information, find a place to park, verify that you are going in the right direction, check on the weather conditions and tune in to a Twitter feed, entirely by means of voice input and responses, so that you can keep your eyes firmly on the road.

A cheeky robot?

Aside from its purely utilitarian function, however, what really makes this virtual assistant unique is that its creator, Ilya Eckstein, set out to equip Robin to be a stimulating interlocutor, with its own personality. Some experts believe this feature will prove a decisive factor in ensuring the popularity – and mass adoption – of VA technology.

During the Virtual Assistant Summit run by events company RE.WORK in San Francisco on 28-29 January, the Robin Labs co-founder shared an astonishing statistic with the audience: among the tens of thousands of conversations that have taken place between Robin and its users, the word ‘love’ figures among the ten most used. If this comes as a surprise when we are talking about interacting with a robot, it nevertheless seems to be a fact that once users realise that an artificial intelligence system is capable of carrying on a conversation like a human being they are keen to humanise it. ‟It’s no doubt because Robin has his own personality and is sometimes impertinent and sarcastic that he appeals to many users,” argued Dr Eckstein.

Social adaptability

However, while giving your virtual assistant its own personality strengthens the affection felt by some users, it might turn others off. After all, not everyone likes the same brand of humour and some people would prefer their robot to confine itself to performing well-defined tasks and not make an effort to converse. This is precisely why it is vital that a VA should be able to interpret a user’s reactions and understand his/her personality and needs in order to adapt better to them, underlines Ilya Eckstein.

Accordingly, Robin has been ‘taught’ to sort its users into a number of categories according to their attitude and their reactions to its suggestions. The robot places users on its ‘Classification by Intent Genome’ chart, with one axis going from ‘chatty’ to ‘restrained’, and the other from ‘communication-driven’ to ‘task-driven’. Depending on where the user ends up being positioned on the graph, the VA may variously take, at one extreme, a strictly professional attitude, focusing on the tasks in hand with no unnecessary chatter or, at the other extreme, opt for a friendlier, chatty style, with a possible range of behaviours in between the two.

So it now seems to be feasible to create artificial intelligence that has its own ‘personality’ but is also able to adapt flexibly to a range of user profiles and find an approach that best suits any given user. Its basic character may remain the same, but its way of expressing itself, the amount of words it uses, its linguistic ‘register’ and tone and the suggestions it puts forward will all vary according to the person with whom it is talking – just like human beings learn to do, basically.

Legal mentions © L’Atelier BNP Paribas