Social Technologies
Authors: Negin M Harandi– PhD student, Department of Sociology, Maynooth University. Funded by ADVANCE CRT (SFI), Fatima Ayoub – PhD student, Department of Electronic Engineering, Maynooth University. Funded by ADVANCE CRT (SFI)
Can robots be social? Can we start to think of them as our co-workers, companions, teachers, or caregivers? Would they make our lives easier or is there a robot-dominated dystopian future already awaiting us? These are some of the questions that might pop into anyone’s head when hearing about social robots. Thinking about how AI could be more social is always interesting. We are an extremely social species and often like to imagine what happens when something other than a human is able to act like a human.
On 19 April 2021, the ALL institute hosted a seminar on “Social robots: the ultimate test for AI and robotics”. Professor Tony Belpaeme (Ghent University) introduced by Dr Rudi Villing (Maynooth University), took us through a few experiments with social robots in different social settings and we got the chance to see how social robots were performing as teaching assistants, therapy assistants and caregivers.
This was especially interesting to us as our PhD projects are about robots. Although we have different ways of studying and looking at social robots as a sociology student and an electronic engineering student, being able to see some of the shortcomings and challenges of using social robots in real social environments was fascinating for both of us.
So, what is a social robot? While traditional physical robots only interact with the physical environments (such as vacuum cleaner robots), social robots are designed to interact with humans and social environments.
The key features of a social robot are that it interacts with people in a socially acceptable fashion and conveys intention in a human perceptible way. To do this, a social robot embodies the combination of a mechatronic interface, artificial intelligence, and understanding of the surrounding human environment. Social robots are designed to engage with humans and effectively provide services to users. They could be designed to perform different tasks such as teaching, therapy, taking care of the elderly.
One of the most interesting parts of the talk was about the experiment with a social robot as a teaching assistant. Before going through what happened in the experiment, it is interesting to know why some of the roboticists believe that using social robots could be beneficial in education. Education is changing and nowadays there is more need for personalised learning and customising education for each student’s needs and skills. It is believed that social robots could potentially bring more flexibility into the educational system.
However, Professor Belpaeme showed us how a more socially capable robot is not always the key to success and better results. Professor Belpaeme and his team observed that a teacher robot with less social interactions was more effective in comparison with a social robot that was making eye contact, using children’s first names, and waited for them to finish the activity before moving on to the next activity.
This was a surprising result for the team. It is likely that children considered the social robot more like a friend and also enjoyed looking at it which made it difficult for them to focus. This is an example of the difference between expectations in the design process and what robotic technology could really bring into our lives. Just imagine being at pre-school again and having this cool, funny, and kind robot as your teacher. It is fun but It would make it harder to focus on your studies, right?
Speech recognition is also a challenging domain in the development of social robots. Social robots need to understand what humans are saying in order to communicate with them. There are two communication manners: verbal (speech dialogues) and non-verbal (gesture, body posture and movements), both requiring artificial intelligence and machine learning techniques for implementations.
Professor Belpaeme talked about a few speech recognition engines (Alexa, Watson, Google assistant) which are developed by well-known companies but still have some uncertainty.
The quality of automation is also a key challenge in social robots. For example, the physical parts should synchronize with the speech dialogues of the robot. On the other hand, machine learning is hard to apply for nonverbal communication because this area is still in need of more focused research and development.
Speech recognition becomes more problematic in the areas where social robots are interacting with the elderly, children, or people with different accents. Therefore, there is a need for a more inclusive design of social robots as current social robots might end up excluding certain groups in society.
In the context of assisted living, social robots have proven to be beneficial as long as they are being used in a purpose-built environment. But many social environments are not yet robot friendly. For example, social robots require full Wi-Fi coverage to function. At the same time, many aged-care facilities have inadequate Wi-Fi coverage. Another example is the robot which was designed to shop for groceries for the elderly population and could only perform in places with wheelchair-accessible sidewalks. These examples highlight the need for designing robots according to different social environments and the user’s capabilities. This means avoiding one-size-fits all designs.
Mylo , a robot developed by CR robotics and its founder Candace Laufer, was also virtually present in the seminar and we got a chance to meet the cute cat-robot. This was a great opportunity to learn more about what a socially assistive robot like Mylo is capable of doing in real-life settings. Mylo is designed to assist people with dementia and their families (https://www.heymylo.ie/benefits). It provides users with more safety (fall responses-heart rate response- remote monitoring- guard), independence (medication prompts- hydration prompts- daily living prompts- appointments/visits) and connection (video calling- entertainment- social pushes- companionship).
Of course, the use of social robots in real social environments raises many ethical questions. The ethical concerns were brought up in the discussion (the panellists: Dr Rudi Villing, Prof David Prendergast, Dr Linzi Ryan, Prof Tony Belpaeme and Candace Lafleur). For example, one of the concerns is that social robots might lead to loss of human contact for the elderly population. The panellists believed that social robots could, in fact, help the elderly population feel less lonely.
For example, robot ‘Mylo’ is capable of making it easier for users to make video calls as well as reminding them to involve in human interactions. Additionally, some areas of care require less human contact (such as assisting the elderly with using the toilet) and robots could be considered beneficial in these areas.
If we look at the social isolation of the elderly population from a sociological perspective, studies suggest [AC1] that human caregivers should never be completely replaced by social robots, even if it is technically and financially possible to do so. This is because humans are always in need of inter-personal interactions and introducing robots into elderly care settings should not make the elderly feel lonelier.
To avoid social isolation of the elderly population, there is a need for more research as well as more rules and regulations about care robots. For example, transparent rules and regulations about the limitations of using robots in elderly care settings, the types of care activities that could be done by robots, the need for human carer’s supervision, the robot’s autonomy, etc could be beneficial (some of these are mentioned in the draft of rules and actions for excellence and trust in Artificial Intelligence).
Some of the shortcomings and ethical implications could be addressed by more collaborations between industry and a broader range of academic disciplines, including social sciences. However, as the panellists pointed out, academic research moves at a slower pace compared to industry which focuses on delivering products as soon as possible. This makes it difficult for industry to take academic research into account when designing and manufacturing robots.
On a final note, designing and building autonomous social robots is a very challenging goal. The main thing we learnt from the seminar is that social robots are still very far away from what we understand as human sociality (e.g., in terms of emotions, empathy, touch, non-verbal communication). Also, our knowledge of their capabilities and shortcomings are still limited as they have not been tested in many real social environments.