Laying the groundwork
The concept of avatars, digital representations of individuals, has existed for some time. Still, the development of AI-generated avatars has taken this idea to a new level. The early history of AI-generated avatars can be traced to the 1970s and 1980s, when computer scientists and researchers began experimenting with early computer graphics and artificial intelligence forms.
One of the earliest known examples is 3D Art Graphics, a set of 3D computer graphics effects written by Kazumasa Miyazawa and released in June 1978 for the Apple II. Early attempts at creating virtual characters were rudimentary compared to today’s standards, but they laid the groundwork for the future development of AI-generated avatars.
Ahead of its time
As computing power increased and AI algorithms became more sophisticated, the capabilities of AI-generated avatars also expanded. In the late 1990s and early 2000s, we saw the emergence of AI-powered virtual assistants and chatbots, representing an early form of AI-generated avatars. These virtual characters could interact with users in a limited capacity and perform basic tasks.
Microsoft’s Clippy or Clippit was an early example of these virtual characters. It was an interactive virtual assistant in the form of an animated paperclip designed to help users navigate Microsoft Word. According to Alan Cooper, the ‘Father of Visual Basic’, the concept of Clippit was based on a ‘tragic misunderstanding’ of research conducted at Stanford University, showing that the same part of the brain in use while using a mouse or keyboard was also responsible for emotional reactions while interacting with other human beings, thus is the reason people yell at their computer monitors.
Microsoft concluded that if humans reacted to computers as they responded to other humans, including a human-like face in their software would be beneficial. As people already relate to computers as directly as they do to humans, the added human-like face emerged as an annoying interloper distracting the user from the primary conversation. As an iconic symbol of its decade, Clippy was also famously terrible and may have been a generation or so ahead of its time. Worldwide consensus decided that Clippy was annoying and intrusive. Time Magazine even named it among the 50 worst inventions ever.
Picking up pace
One of the significant milestones in the history of AI-generated avatars was the development of deep learning and neural network technologies. These advancements allowed the generation of more lifelike and realistic avatars that could understand and respond to human speech and gestures.
People use platforms like Elai and Synthesia for AI-generated avatars in various industries and settings. The possibilities are endless, from virtual influencers and digital celebrities to interactive customer service representatives and AI-powered teachers. As AI technology continues to evolve, we can expect to see even more advanced and realistic AI-generated avatars in the future, blurring the lines between the virtual and physical worlds. One of the most interesting potential use cases for AI-generated avatars is in older people’s and dementia care.
Application to care
According to the Organisation for Economic Co-operation and Development (OECD), dementia is on the rise around the world. In the 38 OECD member states, there were 15.7 cases of dementia per 1,000 inhabitants in 2021 – but by 2050, almost 30 in every 1,000 people are expected to be living with dementia.
NTT Data, a Japanese multinational information technology (IT) service and consulting company headquartered in Tokyo, is testing whether avatars can stimulate dementia patients. In addition, Thomas Nørmark, the Global Head of Innovation at NTT DATA Business Solutions, wrote about the dilemma facing his family when his grandmother, who battled dementia, asked about her deceased husband every morning. These experiences led Thomas to think about how carers can bring back the memory of the deceased.
The Living Memory project, a collaboration between NTT Data and Massachusetts Institute of Technology (MIT), was created to improve the quality of life of dementia patients in their final years and stimulate their memory to slow the progress of their symptoms. This technical breakthrough means people with dementia can communicate with loved ones using a tablet or computer.
For example, the avatar of a deceased loved one knows birthdays, anniversaries, shared vacations, pets and other family members. These facts are implemented in the knowledge base along with their interests, like gardening, hiking and cycling, and this information allows the AI to initiate and hold a conversation. The more information in the knowledge base, the easier it is to train a personality and create a virtual character.
Improving AI’s accessibility
I started GG Care to improve the lives of people with dementia and their carers with readily available smart home technology, after seeing Amazon Alexa’s positive effect on my grandmother. GG Care’s Alexa-powered virtual dementia care companion simplifies everyday tasks through Interactive Reminders (IRs). Carers can use GG Care’s web interface and Alexa Skill to provide interactive reminders for their loved ones.
The IRs simplify the complexity of tasks by breaking each task into easily digestible steps with a series of questions, prompts and hints, just like a caregiver would. Many consider an Amazon Echo the speaker-device avatar for Amazon’s Alexa AI software. The care recipient interacts only with their voice; there are no clunky menus or buttons. The wide range of IRs include reminders to eat, stay socially connected, take medication and more.
I received grant funding from Innovate UK to create an avatar, cloning my likeness to produce videos about technology and dementia care. One of the use cases GG Care is exploring is cloning the likeness and/or voice of dementia carers to deliver IRs virtually, increasing a care recipient’s adherence to the instructions being asked of them. IRs delivered by an AI-generated avatar could remind the care recipient to complete tasks to remain independent, decrease feelings of isolation by engaging in dialogue and promote Cognitive Rehabilitation Therapy (CRT). There is not much information on the efficacy of avatars in dementia care, but such studies are already underway.
Additional research
According to Kerry Yu Fang, a researcher at Western Sydney University, ‘An initial study is currently underway to test the effectiveness of an avatar-based reminder application to improve adherence and overall satisfaction in users. The application is tested on subjects using supplements as a proof-of-concept study. Results from this study should give an indication of the effectiveness and feasibility of using interactive avatar technologies for improving adherence in patients.’
Secondly, Margherita Rampioni of the National Institute of Health and Science on Aging, has argued that the field of Embodied Conversational Agents (ECAs) is novel and poorly discussed in the scientific community. As such, more sophisticated study designs and proof of the approach’s efficacy are required. However, research published by Eduardo Carrasco of Viacomtech found all of his study’s 21 users with mild to moderate Alzheimer’s engaged naturally with an avatar, understood the information conveyed by the avatar and answered successfully using a television remote control.
Possible limitations
Although AI-generated avatars for people living with dementia can offer many benefits, there are also several potential limitations and challenges to consider:
- Technical issues – like all advanced technology, AI-generated avatars sometimes have glitches, hallucinations or malfunctions, which can disrupt the continuity of care and frustrate care recipients and their caregivers.
- Limited capabilities – although AI avatars have made and continue to make significant progress, their ability to understand and respond to complex emotional cues or nuanced situations still needs to be improved.
- User acceptance – people with dementia and their carers are usually older adults. Older adults have higher negative preconceptions around new and unfamiliar technologies, which can hinder their adoption.
- Privacy and security – AI systems often require access to personal data to function effectively. Ensuring the privacy and security of this data is a significant challenge, especially given the sensitive nature of health information.
Ethics are among the most essential considerations technologists and caregivers face when deploying AI. AI in caregiving might reduce human contact and make the care recipient overly reliant on machines for emotional support. With AI-focused research generating plenty of pause for thought, the Scottish Dementia Working Group (SDWG) met to discuss AI approaches in dementia research and whether they could support the NHS more broadly.
Feedback from SDWG’s members suggests they favour AI taking on more laborious and administrative tasks that can take up considerable time for medics and carers. However, the overwhelming feeling among the group is that AI should never, and could never, replace human-to-human interactions. There was also concern that avatars could symbolise a degree of deception – because they give the impression of a real person being present, complete with a representation of their face, voice, expression and pre-programmed knowledge. Essentially, it may not always be possible for people with dementia to distinguish reality from perception. These limitations must be explored further and overcome before the sector’s widespread acceptance of AI.
Are AI-generated avatars the future of caring for people living with dementia? Should we approach this new field with caution, if at all? Leave a comment on this feature or join the conversation to share your thoughts.
David Grey is the Founder of GG Care.
Email: [email protected]
X: @GGCareApp