Stevie the Robot. Image: Akara Robotics
Niamh Donnelly has a crucial role to play in looking after ageing populations as she works on the ground developing AI for Stevie the robot.
The world as we know it is ageing. Globally, life expectancy increased by more than five years between 2000 and 2016 – the fastest increase since the 1960s. In Ireland, where our youthful population has long been an attractive selling point for foreign direct investment, the script will begin to flip by the close of this decade.
According to official projections, we’ll begin to see the ratio of younger to older people in Ireland shift in favour of the latter by 2031 and continue in that direction. By 2051, we expect between 496,600 and 691,800 more older people than younger people in the country’s population.
An ageing population means increased demands for elder care, which can range from extensive medical needs to day-to-day assistance. A vision of robotic helpers in this space is no longer seen as a futuristic remove as it becomes a present-day reality.
Currently, the social robots that are beyond trials and prototyping are little more than simplistic companion devices, such as Paro, the fluffy, therapeutic robot seal. But work is ongoing to advance social robotics to offer more functions and interactions, and Niamh Donnelly is helping bring this idea to life.
Suddenly, Stevie
Fresh from securing a master’s in computer science from University College Dublin and winning an AI Award for a student project in 2018, Donnelly joined the Trinity College Dublin team behind Stevie the robot.
Stevie – who will be appearing at Future Human alongside Dr Conor McGinn – is a Time magazine cover star lauded as “the robot that could change the senior care industry”. Donnelly joined the team just as they were spinning out as Akara Robotics and was tasked with leading the development of Stevie’s artificial intelligence (AI).
“My area of focus is on improving Stevie’s ability to communicate and interact with people autonomously in social settings, such as in the lobby of the nursing home or during a game of bingo when the robot is surrounded by people,” she explained.
“This is a really hard problem. While humans naturally pick up on social cues and group behaviours, this sort of thing is much more difficult for a robot. Even simply answering a question involves Stevie to recognise who a person is, if the question is directed to Stevie, where they are in the robot’s field of view, what the question means, what an appropriate response might be etc. And we expect the robot to all of this in an instant.”
To achieve this, Akara has run trials using heavy algorithms, computer vision, machine learning and dialogue systems, putting Stevie’s computing power to the test. “We’re focusing now on low-power AI-on-the-edge devices,” said Donnelly. “It is challenging but exciting work.”
Image: Niamh Donnelly
Field research
Donnelly herself was featured in Stevie’s Time cover feature as she was embedded in the field at Knollwood Military Retirement Community in Washington, DC. She and her team spent weeks on the ground observing how Stevie and the Knollwood residents interacted, and noting any necessary improvements they needed to bring to these experiences.
“In Akara, we like to have engineers on the ground,” explained Donnelly. This, she said, isn’t a common enough element of the development of social robotics just yet, but has been essential for Akara to discover just how Stevie should work.
“The success of our AI programme is not just dependent on what software tools we use. Nursing homes and retirement communities are quite unique places, and having now spent so much time there, it has helped me get a better understanding of how this all might come together and work in practice.”
In its latest iteration, Stevie has a dialogue system that can work in specific contexts. But during the pilot study, Donnelly took control of the robot for other scenarios not yet programmed.
The study proved hugely valuable to the team, not least because the elderly population is “brutally honest”, according to Donnelly. “On the ground, feedback is immediate,” she said. “They were not shy!”
Something Donnelly understands from this feedback is that humans don’t necessarily want humanoid robots. Though the design and mechanics of the future will enable increasingly lifelike forms, fears of getting lost in uncanny valley abound. “The feedback is that Stevie works, so we’ll stick with the current form,” she said.
While decidedly robotic looking, Stevie has been designed with elements to delight those he interacts with. “We put a lot of thought into his expressions,” said Donnelly, which include sadness, happiness and surprise.
Even when something went wrong, the result was surprisingly helpful. “We were worried about mistakes in the beginning, but that made him more human,” said Donnelly. “People warmed to him more.”
The future of elder care
As development of Stevie continues, Donnelly hopes to see an acceleration of products like it from research. But that’s not to say she thinks all elderly people should be abandoned to robotic helpers in the future.
“On a personal note, we should be sceptical of AI and machine learning introductions and repercussions,” she said. “We were nervous with the Stevie project of the reaction from care workers, but they welcomed it as useful.”
It’s this vision of social robots enabling carers to interact more with their charges that Donnelly sees for Stevie and other elder care innovations. For example, because Stevie can take over the job of running a game of bingo, calling numbers and listening for winners, carers are free to work the room, interacting with the residents.
“Carers can be preoccupied with running an activity, but Stevie frees them up for personal care. Elder care then benefits from more human-to-human interaction,” said Donnelly.
‘We were nervous of the reaction from care workers, but they welcomed it as useful’
– NIAMH DONNELLY
Right now, the plan is to continue researching and improving Stevie with the residents and carers at Knollwood, and Donnelly credited the nursing home for investing in the forefront of technology for its residents. “They’re the perfect partner,” she said.
She would also like to continue testing with nursing homes in other countries and tease out local, cultural differences. One study group wanted Stevie to dance, while that’s not something they asked for in another study. “It shows that people have difference desires. I would like to explore more the differences between groups.”
Akara is actively pursuing opportunities for Stevie in Europe following a successful pilot in the UK last year. “We have been collaborating with Alone since 2017, and following this we have been in active discussions with several organisations in Ireland for the past several months, and are hopeful that we’ll have confirmation of a first Irish test site in the near future,” added Donnelly.
“We see huge opportunity in Ireland, both as a base for our R&D operations as well as deployments of Stevie, where we are confident it will help lead to improved care outcomes and working conditions, which I’m sure you’ll agree our country desperately needs.”
Future Human is Silicon Republic’s international technology, science and business event celebrating meaning, values and purpose with collaborative, interactive and hands-on experiences. Early Bird tickets for Future Human 2020 are available now.
Want stories like this and more direct to your inbox? Sign up for Tech Trends, Silicon Republic’s weekly digest of need-to-know tech news.
Editor’s note: Please note that Future Human 2020 has been rescheduled to take place on 29 and 30 October 2020. Speaker line-up subject to change. More details can be found here.
