Building Trust Through Q&A
Uncertainty is the driver of research. To find out what we don’t know, to build our reservoirs of knowledge and to learn who, where, what, when, why and, most importantly, how? – we ask questions.
Autonomous systems are becoming an everyday part of more and more diverse industries. It is important that the people working with them find them trustworthy and reliable, so as to have confidence in the outcomes.
Senior Researcher Damien Teney has been working with Dr Qi Wu and Dr Ehsan Ebbasnejad on developing trusted autonomous systems that function reliably in a complex environment and work collaboratively with people to solve problems.
To do this the autonomous systems need to be aware of uncertainty, ask questions to resolve that uncertainty and provide clear and explainable decisions in response.
This would mean they understand their interaction with the world and are capable of applying reason to a situation.
The primary technology developed in this field has been in what is called visual dialogue, which enables an ongoing conversation about an image or other visual information.
Damien says his team’s research enables questions to be asked proactively, so it’s not always left to people to drive a conversation and the autonomous system to just respond.
“This can fill in the ‘unknown unkowns’ by supplying questions we may not have realised needed asking and therefore help to find answers we didn’t necessarily know we were looking for.”
These conversations aim for a specific outcome, cutting out pointless chat and speeding up a result.
It is a significant step towards achieving the goal of real active collaboration between people and the autonomous systems they work with.
Story provided by Dr Damien Teney.