Graduating with iCat – Research in the perception of social intelligence in robots.

Home / Blog / Graduating with iCat – Research in the perception of social intelligence in robots.
Graduating with iCat – Research in the perception of social intelligence in robots.

I thought it would be fun to write a blog post about my master’s thesis and research project. (Amsterdam University, 2006). I had so much fun doing this project and it really laid the foundation of my UX career and my Robotvilla endeavours.

What was my research about?

In my research project, I studied the perception of social intelligence in robots. The what of what?? Well, in other words: Do people perceive robots to have social intelligence if the robots use facial expressions while performing a task in a certain context? And if that was the case, how do we respond? We were privileged to have the Philips iCat for our study. This yellow cat-like robot was able to display different facial expressions, play sound, record the person sitting in front of her by using a camera in her nose, respond to touch, and enhance her emotional state by flashing her multicolour LEDs. iCat was created by Philips especially for research on this topic and there were very few of them. So, so, so cool! I wish I could get my hands on one for my Robotvilla. Let me know if you can help, I would be forever grateful!

Different facial expressions.

The research

In a nutshell: We found sixty participants who were willing to perform a task on an EBay-like auction website. They were asked to search for and bid on a few items using only an iCat and a computer screen. So, no keyboard and no mouse. I created two versions of the iCat: a social intelligent version and a neutral version. The social intelligent version displayed facial expressions and had an expressive voice while interacting with the participants, the neutral version didn’t have the facial expressions and spoke in a monotone voice. The assumption was that performing the tasks using the social iCat would be more successful and enjoyable than with the neutral iCat.

To execute the tests, we used the Wizard of Oz method. This way we could pretend that the iCat worked autonomously, while in fact I controlled the iCat while I was sitting in a separate room nearby. If you are interested in the outcome of my research, you can read my master thesis right here. (It’s in Dutch, but if you are interested in an English translation, please send me an email.)

This setup was what the participant had to work with. No keyboard, no mouse, only an iCat.

Highlights in my graduation project.


Two other students were also using iCat for their master thesis, so we joined forces. This work method had of course pros and cons… no wait… there were only pros, because we were a dream team! We really hit it off. The benefit of our team was that we completed each other. For example: I found the data analysis part a bit difficult and I wasn’t used to the SPSS data analysis software. One of the guys helped me out with that. On the flip side, I helped them out with programming the iCat, coding the fake auction site and creating the research scenarios they needed to test their hypothesis. Win-win!

(Video: This is what happens when you have to run a research scenario 60 times – over and over and over again… We decided to have some fun with iCat *hihi*! Look how young we are (2006)!!)

Research design and  programming the iCat

The participants had to navigate around on the auction website and search for different items like a snowboard, a phone, and a furby. To do that, the participants had to interact with iCat verbally. We created scenarios with all the possible answers and reactions iCat could give in a certain situation. Next step was to program the facial expressions of iCat and create the Loquendo text-to-speech sound files that fitted the reactions iCat could give in the iCat OPPR suite using the LUA scripting language. This was so much fun to do! In the picture below you can see the reaction-panel I created for iCat, with columns for the different steps in the scenarios. I grouped all the reactions in different categories. The categories were ‘introduction’, ‘search’, ‘find’, ‘not found’, ‘bidding overruled’, ‘the end’ and ‘general’

iCat interface using LUA
Wizard of Oz – iCat reaction-panel

Programming in LUA and the OPPR software was so much fun and thinking back, it fuels my hunger to pick up coding again. On this topic I wrote another blog post. It was quite a challenge to sync iCat’s movements with the sound files I produced using the text-to-speech software. In the video below you can hear the voice I created with coughs and all, and you can also see a big part of the bidding scenario.

(Kind of a long video, but you get the idea :))

Final thoughts

In this research, I used a lot of techniques I still use today while I practice UX design: prototyping, user journeys, scenarios, usability testing, questionnaires, observation techniques, data analyses and reporting techniques. My graduation project really feels like the start of my UX journey and my endeavours. I feel grateful for the opportunities I have received.


Can you help me in my quest for an iCat. I would really, really, REALLY like to introduce her to the kids in the Robotvilla.