You stole with a robot, now what?

Last summer we accidentally stole something with a robot” seems more like the tag-line for a cheesy science fiction movie than a starting point for an academic paper. Yet it was one of the events that we describe in our latest paper titled “Telepresence Robots in the Wide Wild World” which is accepted at the CHI 2017 Ethical Encounters workshop. Congratulations to my co-authors: Robby van Delden and Jered Vroon.

Right, let’s first look at the “crime” and then I’ll explain what really happened. 🙂

In the paper we describe a method that can help to find societal issues with new technology. The Researcher As Experimental Subject (RAES) method basically means that as a researcher you take the new technology with you, out into the wide wild world, as often as possible. You run into many real-world situations that are difficult to predict from a comfortable arm chair and that might not occur at all in a laboratory setting. This is the main value of the RAES method.

Now, as with any scientific research we had to get ethical approval. This is to make sure we don’t do anything that is wrong. Within our department we can get ethical approval for, what is called, standard HCI research. That approval covers most research where nothing “interesting” or difficult happens. Examples of this standard research are: do you like better website A or B, how would you describe the personality of this virtual human/robot, which of these projections seems sweeter to you, or would you help this guy with his telepresence robot? Not something where you would expect anything to possibily go wrong or hurt someone. As our experiment seemed ethically sound and safe, we got approval to conduct our RAES experiments.

We took a mobile telepresence robot out into the real-world. We it took all the time and everywhere, for a nice overview see the project website (*Bot), but most notable was shopping with the robot. This went surprisingly well as you can see from this clip:

Cookies acquired! Time for some drinks:

Great stuff, but it did bring up questions about legislation and identification. Could the cashier know my ID was not fake? What if my 15 year old niece used a recording of my “BeerBot” interaction to buy herself a beer? Should we have a license plate for these telepresence robots so the owner can be tracked? This shows the benefits of the RAES method, since we can now easily see where challenges and opportunities for this technology lie in this situation.

The theft

Taking the issues of identification a step further, we tried to find people willing to become accomplices of theft, as you could see in the first video. To our surprise this was not difficult! The cashier, well informed about the project and her roll at this point, was distracted by the recruited accomplices in order to let the robot pass by unobserved and without paying. Free beer for all! No we’re joking, it was returned (obviously).

The willingness of the participants to help us steal something surprised us. If we would have to guess, this might be due to the novelty of the technology. Confronted with something new, people need to spend some amount of cognitive power (thinking) to deal with the new situation. He or she can not rely on automatic pilot in a new situation. This is normally a good thing, it helps us adapt and survive in new and unexpected situations. But, the brain can only handle a limited amount of cognitive load! We hypothesize this means that the participants in our study did not have enough “brain power” left to dismiss our request! Realising this, during the experiment, we conducted an extra sensitive debriefing of the participants. In a debrief we explain, amongst other things, why we did what we did and in this case that stealing is not appropriate behaviour.

This case also raises the point, who is responsible for a theft with a telepresence robot? Is it always the owner of the robot? But what if his robot gets hijacked or drives out of a shop without paying because the connection went bad? In addition, we need to investigate the ethical issues with doing RAES research. The ‘theft’ serves as an intriguing example of how to deal with ethical approval for ad-hoc experiments: “things you think of in the field”.

In our paper we discuss ways to ensure that research stays ethical with methods such as the RAES method. One way to go about this, is give researchers ethical training. This allows them to make ethically sound decisions in the field. In addition, we proposed to never go past n=1. This basically means that you try it only once and then go to a quiet corner to reflect. Or better yet, discuss the case with your ethical expert. If this can’t be done before the experiment, than at least do it after n=1. Research in the wide wild world can give us great insights, but it has to be done properly.