Personality & robots. A matter of psychology or statistics?
2018 is a break through year for Voice. Smart speakers for example have entered over 30 Million homes. Both tech firms and brands are starting to heavily invest in this new platform. Dr Fiona Chan researcher at the psychometric centre of Cambridge University, took me on a deep dive as to where we stand in making chat bots personality sensitive and responsive.
Time-sequenced questions and transcript of the interview below.
- 00M08 Fiona Chan, you’re a researcher here at the University of Cambridge. Can you introduce yourself?
- 00M50 Could you tell something about what your research question actually was and what you discovered along the way?
- 01M49 The research is continuing, could you please elaborate a little on how you think the research will be conducted in the near future?
- 02M36 Is there anything that you can more or less predict that might turn out to be the case in robots and personality?
- 03M41 So it will be more statistics than anything else, do you believe?
- 03M59 As a psychologist, do you find that encouraging or disappointing?
- 05M01 Creativity, empathy and improvisation, many people believe are skills humans are beating computers at. How long, do you think that statement will hold?
- 07M15 If computers become so advanced in creative thing, empathy and improvisation, how efficient could a robot be at playing poker?
- 08M40 Having done all this research in robots and personality, can you distinguish some trends in the directions of research in this area?
- 10M05 Are we still trying to find out how to make robots personality sensitive? Or are we already entering the phase of becoming adaptive to the personality?
- 11M19 Will we see in the near future romantic relationships between people and their robots?
- 12M20 So, we as humanity don’t need to fear robots in the near future?
00M08 Fiona Chan, you’re a researcher here at the University of Cambridge. Can you introduce yourself?
Hi, I am Fiona Chan. I had an internship in Microsoft Asia research in 2016, when I researched on how to make chat robots more psychology consistent.
For instance, how to make a chat robot have a more consistent personality, and also how to allow chat robots to understand the personality of users.
00M50 Could you tell something about what your research question actually was and what you discovered along the way?
My research question was to see whether we can develop statements or items that can be used by a chat robot to reflect personality traits of users.
For example, if want to know if my user is an introvert, I will start the conversation, or I will ask my users whether they have plans for dinner at night, or whether they have plans at the weekend.
Something like that to reflect or to understand whether the person is extroverted.
01M49 The research is continuing, could you please elaborate a little on how you think the research will be conducted in the near future?
We have already developed items or some statements that reflect the personality of the users. They are all in forms of daily life examples.
So, what we want to do in the future is automate, to make the script using these questions and automate the process of data collection. Because this study will need massive data to reach findings.
02M36 Is there anything that you can more or less predict that might turn out to be the case in robots and personality?
I tried to examine the quality of each statement that I made or created. However, I think in the future perhaps there is not much need to focus on psychology in creating a good chat bot that makes your users engaged.
Basically you can use statistics to calculate what you users want to say, and what they want to know. You don’t have to interpret this information with psychology or personality. So this is what I believe in the future, they might exclude psychology and make everything a black box.
03M41 So it will be more statistics than anything else, do you believe?
Yes that’s what I believe. Because business or companies want to make a product that works, instead of understanding what’s behind.
03M59 As a psychologist, do you find that encouraging or disappointing?
I find that both encouraging and disappointing. It’s encouraging because it does something that we could not do. It predicts what the person wants.
Even with our theories and all this research, we couldn’t reach that point where we can make such good predictions of what our users want. However, with statistics they made it.
But it’s discouraging because the area, psychology, seems not to be important anymore because they are not making products as much as statistics does.
05M01 Creativity, empathy and improvisation, many people believe are skills humans are beating computers at. How long, do you think that statement will hold?
I think we are already beaten by AI in these few areas. For instance creativity, an acquaintance of mine in Microsoft research had developed a robot that could write very beautiful poems. You can upload any pictures to the robot, and the robot will write a poem that is about that picture. The poems are sound. They are elegantly written.
So, I think creativity is not a problem with AI as long as there are rules to follow. For example music, to create good piece of music– musicians first have to learn the theories behind notes. For instance they learn that what interval would make perfect harmony.
What notes cannot go together because it just sounds terrible. So, they learn all these rules and they try to compose and make beautiful music. With these rules AI can also adapt. It won’t surprise us if a robot can write a piece of music as good as a human does.
However, I do believe humans and robots are not in competition. First, we developed these robots, secondly is not necessary to compete. Just appreciate each other’s skills. They are good at something doesn’t mean they are better than us. We can be equally good.
07M15 If computers become so advanced in creative thing, empathy and improvisation, how efficient could a robot be at playing poker?
Yes, possibly because on a poker table you try to guess the person’s, the competitor’s intention. There is a lot of bluffing going on. There are lot of conversations and guessing going on.
I think from this point of view a robot can play really well on a poker table, given that the robot can predict the intention or its competitors.
I think the real spirits of poker, or the real spirits in poker is to be inconsistent so that your competitor can not guess, or cannot expect your actions.
So then from this point of view it is actually quite difficult for a poker robot to gather consistent data, and to make meaningful predictions.
Perhaps if the data is large enough for a robot to make a prediction that is okay. But I don’t think we can reach that amount of data for each person just on the poker table.
08M40 Having done all this research in robots and personality, can you distinguish some trends in the directions of research in this area?
It depends on how you define robot. If it’s a machine and it is not interactive, then there are many areas that try to predict users or people’s personality with algorithms.
For instance, digital footprints, internet of things. In the field of chat robots there are two very different strings. They sound very similar but are actually very different. One direction is to make the chat robot more consistent in personality.
Another direction is to help or devise a robot that understands the personality of its users. In the hope that in the future the robot can use this information and adapt to this information and make the users more engaged.
10M05 Are we still trying to find out how to make robots personality sensitive? Or are we already entering the phase of becoming adaptive to the personality?
To my knowledge, most of the research projects focus on making robots personality sensitive. So there are many different attempts, they attempted from different directions. Well it’s actually not very difficult to do if it is a structured conversation.
However the tricky part could be understanding someone’s personality in a casual conversation. If the conversation is very structured, it is actually very easy to get information. But if you want to understand your user like a friend, then it is more difficult.
11M19 Will we see in the near future romantic relationships between people and their robots?
It depends on how intelligent the robot is. But if the robot is pre-programmed then I don’t think there is a lot of fun in a romantic relationship. On one hand, we want people to be attached to robots to solve problems.
For example they are lonely, they might need a companion. On the other hand, we don’t want people to get too engaged to robots. They are pre-programed for fantasy, it is just a preprogramed thing. It is not real. You don’t want people to get obsessed in this. It is sort of a dilemma.
12M20 So, we as humanity don’t need to fear robots in the near future?
Yes and no, but I would choose no.
Interview Peter van der Bel, founder & curator of The Centre for Applied Product Personality Research