Karthik Prakash

I found the course to be an interesting sequel to Conversational Agents. The idea of creating a virtual agent and embodying it on a robot that can perform human-like movements is exciting but a little bit terrifying at the same time. The course taught me that even if the robot is built for a section of the population, the needs of users are unique from person to person. For example, people with dementia tend to have different levels of physical and cognitive disabilities and it is imperative for developers like us to make sure that the agents are personalized to each one. This is one of the reasons why creating an agent to help people with dementia was a challenging ordeal for us. 

As we worked on our prototype, I realised that my group members had different opinions on how the ideal robot agent should be. Demi, who comes from a communications background, wanted the robot agent to communicate effectively with all the stakeholders while Arber and Valentijn, focused on the feasibility of the interactions since they come from a Computer Science background. Andrejs focused on making the Flutter application smooth and efficient while Zenan focused on the different interaction scenarios. Being a research aficionado, I did lots of research on the needs of people with dementia and the agents that were created to help people with disabilities. I found some fascinating analogies on how people act with the robot and how the robot should respond to different cues. I helped set up the evaluation environment and worked with Arber on developing the Wizard of Oz for the Pepper robot.

During the evaluation of our robot, we felt a rollercoaster of emotions. After getting 3 people for the first 3 hours, we were thinking about alternatives if we did not get enough participants. But we learnt that it's not over until the chequered flag is shown. Our classmates from the other groups showed up one by one, and along with some of our own friends, we were able to reach the required number of participants for our evaluation. This experience taught me that it is probably not a good idea to send invites the night before the evaluation session.

But, due to the hurried nature of our project, we could not implement all of the features that we wanted and as a result, our robot agent is far from the ideal agent that it should be. But it's a good start. We believe that given enough time and resources, we would be able to build an ideal robot that can help people with dementia and test in a real-world setting.

Overall, this was a great course for learning to build agents that are specifically designed to satisfy the unique needs of users. I learned about the social aspects of robot agents along with aspects of cognition, music and gamification of therapy. I am thankful for the opportunity I received to have worked on a project that can hopefully, make the lives of people with dementia a little easier.