Project Summary
Duration: 6 months
Team: Solo project (my Master thesis)
Partner company: Furhat Robotics AB
Location: Stockholm
Machines are designed to be infallible, but what happens if they are suddenly struck by chronic mental decline such as dementia? For this project, a social robot has been transformed into Max, an elderly mild-stage Alzheimer’s patient. The goal was to evaluate the effects on the users and to derive the potential of using it as a training tool for caregivers and medical students in the future. Through user testing, 20 participants interacted with the robot. The results indicate that the patient simulation was convincing, leading the users to treat the machine as a human being and confirming its potential for educational use.
Problem Statement
At first, the idea of the company was to create a scenario where the robot can identify if its conversation partner is suffering from dementia. Concerned that this would exceed the scope of a Master thesis project and that it would pose an ethical dilemma to receive such a heavy diagnosis from a machine, I decided to switch the tables around. I wanted the robot to become the patient with Alzheimer’s and asked myself:
-
How to simulate an Alzheimer patient with a social robot?
-
Which effects does a social robot simulating Alzheimer’s Disease have on the interacting user?
Solution
The robot persona Max, portraying symptoms of a mild-stage Alzheimer’s patient, was created. It aims to provide an innovative and instructive way for users to experience how it feels to talk to someone with dementia in a safe environment. It conveys the difficulty of the situation, but instead of scaring people away of the limitations dementia imposes on a patient, they should be encouraged to look beyond. They can see that it is still possible to have meaningful conversations and recognize the person behind with a story.
Process
01.
Research about Alzheimer's and the needs of caregivers
02.
Interviews with patient actor and health professionals
03.
Personal meetup with a mild-stage Alzheimer's patient
04.
Creating the robot persona Max and pre-script dialogues
05.
Pilot study with a psychologist and final experiment
Research
The project began with a profound literature research, where I gathered knowledge around the disease and discovered the online database Dementiabank. It offers numerous audio recordings of people having a conversation with Alzheimer patients and gives a direct insight on how symptoms are expressed. As a next step, I scheduled an interview with a patient actor, to learn more about how they typically take on the role of a specific patient. Surprisingly, both her and the training center at the hospital confirmed that none of the actors ever had to portray an Alzheimer’s case.
More interviews followed with two psychologists, a nurse, a doctor and a speech pathologist to reveal more insights about their role and experiences with dementia patients, the needs of caregivers, as well as the potential of a robot simulation. Finally, I had the chance to meet a mild-stage Alzheimer’s patient and it was such a memorable experience, that I decided to translate it in a very similar way onto the users, e.g. I had a pleasant conversation with a communicative person and I almost forgot I had a patient in front of me. Only after 15 min, I noticed she started repeating herself and asked me the same questions repeatedly.
The target user groups are:
Caregivers
As became clear throughout the literature research, caregivers are usually the ones that are overlooked and the original intention was to alleviate their daily lives, as they are the ones that spend the most time with the patient.
Medical professionals
Throughout my interviews I realized that not only relatives but also professionals do not receive a proper training in how to talk effectively to dementia patients and that the demand of learning how to react in certain situations is high.
Anyone interested
Getting a feeling of how it is to talk to someone affected by Alzheimer’s could help to raise the general awareness for the disease and eliminate the stigmas around it.
Persona
Max Henderson, 72 years old
Nationality: British
Family status: divorced, two children, one granddaughter
Occupation: retired English teacher
Hobbies: going for walks, reading
Personality: talkative, joyful, pensive at times
Diagnosis: three years ago, mild stage
Symptoms: word finding problems, forgetting names and numbers, repetitive behavior, mispronunciation
The inspiration for Max’s character traits and life story came from observing interactions with patients online, talking to psychologists and doctors, as well as through a memorable encounter with a real patient. The ambition for the experiment was to create a positive experience. Therefore, the robot was chosen to be in a mild stage of Alzheimer’s and have a pleasant personality.
Experiment setup
The setup for the user: 1. A microphone, 2. The stopwatch, 3. The user's seat, 4. The robot with a kinect, 5. The cheat sheet, 6. Recording camera
The setup for the wizard: I could follow the user's expressions and had prepared buttons for pre-scripted answers of the robot as well as a free text-to-speech field for spontaneous answers
Wizard of Oz setup: For the purpose of this experiment, the robot was not fully automated but externally controlled by a wizard (me). However, the participants were kept with the illusion of a fully automated robot for a better user experience.
Participants: 20 participants were gathered for this experiment. To allow for in-between subject comparisons, they were a mixture of medical and non-medical students, people who knew someone with dementia closely and those who never met any. The experience was evaluated through user interviews along with user observations.
Pilot study and iteration: Due to the reactions in the pilot study, several small changes were implemented. It became clear that the participants would soon run out of discussion topics. This led to the emergence of a cheat sheet with ideas. Additionally, more frequently asked questions were gathered and a need for a standard answer was identified in case the participant is asking a question that would deflect too far from the topic.
The task: The user should talk to the robot for at least 15 min and try to find out details about Max’s life. This task was chosen to see what people are actually interested in, how long they will talk to the robot and how well they estimate its capabilities. It also allows for a loose conversation and reflects the Reminiscence therapy method, to make users understand that a patient can still remember old memories quite well.
Results
The attempt of creating a positive interaction for the users turned out to be a success, as confirmed by the survey outcome, the long conversation times (24 minutes on average) and the many enthusiastic answers in the interview. It could be seen that medical students were part of the group that spoke the longest and asked the most questions to Max, especially regarding his health issues, whereas people who had AD in their close environment were more careful.
In general, participants overlooked the fact that their conversation partner was a robot and reacted very patiently towards Max in spite of his repeating symptoms. They did not correct him, merely helped at times. The way how they talked to him and cared for his feelings revealed that the subjects treated Max as a human being. Some even showed high degrees of empathy for both positive and negative moments and opened up their own lives. In addition, many entertaining situations appeared and all participants would recommend others to undergo the same experience.
A participant is giving Max the time to remember a certain word
A participant is laughing at one of Max's jokes
„At first, it felt a bit artificial. But quickly I began to empathize. Towards the end, I felt some sort of attachment, like in a human relationship“
„Even though he is a robot, I still didn’t feel like asking him about his disease. It seems like a bad thing to do“
Quotes from anonymous study participants
„It conveys the emotional depth that is missing in text books. You get a feeling of the burden for the family“
Learnings
-
The evaluation of the psychologist and the results of the study proved that the robot simulation of a cognitive impaired patient was accurate and convincing the interacting users.
-
To my positive surprise, they show that people are indeed able to empathize with the robotic patient and look beyond the machine, if it is designed accordingly.
-
Since some users even shared intimate details about their family situation, one could imagine how that this “fake” emotional bond could be easily exploited in a manipulative way for information gathering. Therefore it is especially important not to cross the line at any moment and be extra careful with data privacy.
-
This project justified a great potential of using social robots for the education of medical professionals and caregivers, as well as to raise the general awareness for the disease.
Limitations
-
As part of my research process, I tried to get in touch with several AD caregiver communities to be able to talk to them first-hand and understand their needs better. Unfortunately none of them got back to me so I mostly had to rely on secondary research for this part.
-
The user study was conducted with 20 participants, which does not allow for statistically significant conclusions. Also, the group of participants was rather young (average age: 25,7). Older adults in a more suitable age for being a potential caregiver might react differently, e.g. be less patient with a robot.
-
Furthermore, all participants were highly educated, either currently studying of having completed a university degree, which is not representative of the general public. It can also not be guaranteed that participants were mostly acting patient because they knew they were being observed.
Next steps
-
It would be interesting to conduct a follow-up interview with those participants that are in contact with AD patients, to ask them what they still remember from the encounter, if they have applied the same behavior with a real patient and which effects they could notice.
-
Another idea would be to conduct the study on a larger scale, with more variety amongst participants, including actual caregivers, older people and children, and to test an A/B setting, also to evaluate the benefit of using a social robot over a virtual patient.
-
Finally, the next step could be to create a persona in a later stage of Alzheimer’s, expressing more severe symptoms. This would let interlocutors face a more difficult situation, but could train them more effectively for specific situations, such as how to avoid aggressive behavior.
If you are interested to know more about this project, feel free to read my complete Master thesis.