The Wisscomm Project I am a Scientist
I am not too experienced with communicating my research to people outside of the research community (except for, for instance, this event). Of course we do that when we teach, but that’s not accessible to a broader audience.
Recently, I have been contacted by the initiative “I’m a scientist - Get me out of here”. This is a research communication project for school children.
This initative also uses an instagram channel, where they explain more about what they do.
Goals and Procedure
I’m-a-Scientist has, as far as I understand, two goals:
- Give school children the possibility to ask researchers questions and get answers directly from them.
- Show to them that doing research is a possible job choice.
To do that, they have an online platform in which they allow teachers to register and ask researchers to register. Apparently, they do that in rounds where each round is specific to a topic. There was already quite a set of topics in the past, including Social Media and AI and Medicine. I was part of Does AI communicate?. In the list of researchers of this round, one can find a profile of each researcher which describes a bit what they are doing and what they are working on. Mine is here.
Children could then ask questions in two ways: (1) asynchronously and (2) synchronously. The asynchronous approach allowed them to put a question on the platform and some (or all?) researchers on the platform got a notification mail that there is a new question. These could then be answered. I did for instance answer this question.
In addition, there was a synchronous text chat, in which students were put together in a chat system and could ask questions to around 3 researchers who signed up for this time slot of 30 minutes. There was only text and no other modality available. In addition, the live chats had a moderator, and the teacher was also in the room and could write. Except for the researchers, everybody was anonymous.
Own Observations
Overall, I liked this whole thing a lot. I learned a lot what children find interesting and also it was very surprising how different the answers by other researchers (from a similar research field or a different field) were from the own answers.
I’d like to share my impressions in a bit more detail in the following:
General
The general topic was “Does AI communicate”. I don’t know how the organizers come up with these topics, but I liked it. It’s broad enough to attract many people’s attention but still specific enough that, I guess, many people have an idea what this is about. For the case that you read this in a couple of years: we are here a couple of months after ChatGPT and other Instruction-tuned models were made publicly available.
Next to the challenge to find a good name for such a topic-focused round, there were, of course, other challenges to be solved by the organizers. I believe that finding teachers who want to participate with their class is one of them, but I don’t know anything about the process. What I found interesting is the choice of researchers who were involved. I would say they were from very varying fields (see for yourself):
- Researchers who work in communication sciences and want to understand the impact of generative language models on human communication.
- Researchers who study the impact of AI on the society and the future.
- Computer scientists, natural language processing people, computational linguistics, and digital humanities scholars.
- People at the border to art, who want to understand how AI can be used to support creative writing.
I found this selection really good, but interestingly, I was quite often (not always) the only technical person in the chat rooms. However, this did also not really matter: most of the questions were not about the technical background of what makes an “AI”. That’s probably not too suprising: “AI” is, in my perspective, not a concept of interest. It is a buzz word that combines many things, including generative language models, optimization, search, machine learning, or logic. Conflating all of that in one word creates an abstraction that makes it hard to focus on specific aspects. I do not really like that, but I also see that one needs simplified concepts to be able to talk about them.
However, the questions that I have seen show that this abstraction made it sometimes hard to answer them. I do not criticize anybody here – I wouldn’t know how make it better - but I often felt challenged.
Many of the questions that I have seen in the chat platform were about:
- Can an AI take over the world? Can it be dangerous? Can it have a will to survive and replace humans?
- I mostly answered this with “no”, but that it can be used for harmful purposes by people. I typically focused on the aspect that computer systems (including “AIs”) are tools that are used by people. That seems to be something that is not really something that is general knowledge. I knew that this opinion/fear exists, but it seemed to really be wide spread.
- Are you an AI?
- I tried to briefly explain the Turing test.
- How does an AI work?
- I never answered this question because other’s were always quicker, and I was quite happy about this. The explanations by other people were mostly focusing on supervised learning but did not include reinforcement learning, crowdsourcing, or pretraining of models. That would of course also have been totally out of scope.
- Why is it difficult to understand how an AI works?
- I tried to answer that this is not a conceptual issue, but more of a challenge to deal with a complex system.
What I would like to know more is - how have the teachers been prepared, and were they instructed to prepare their students. My impression was that there were huge differences in preparation, both quality and quantity. I think it would be great if the students would know about in advance to whom they are talking - and I’d also like to know more about the students in advance. I will ask the organizers about that and update this post when I receive an answer.
Synchronous Chat
This was an interesting experience, with the synchronous chat. It felt a bit old-fashioned, without emojis, avatars, or direct messages. The threading was also quite basic; perhaps a more hierarchical presentation of the chat would have been a good idea. However, the advantage of a linear chat was that it was easy to understand, and the moderator sometimes paused the questions such that the researchers were able to answer. One feature that I really liked was that, after typing a specific number of characters, the window turned red. A very intuitive way to tell me to keep my answer short.
I was not too happy with the teachers in the chat rooms. One teacher answered in one situation that “even his children” were able to recognize a specific property of an AI. I found that patronizing, and it reminded me what I did not like about school. Another teacher was apologizing for the behaviour of a child, who “tried to be funny”. Come on - what’s wrong with being funny?
Altogether, the chats could have been a bit longer. They were limited to 30 minutes, which was barely enough. There was also no real dialogue. The school children shot their questions and we tried to answer as quickly as we could. I think that smaller groups could have also helped.
Asynchronous Questions
This feature was also used a lot! I, personally, however, did not like it too much. I was always late; when I clicked on a question for which I received an notification, there were already tons of answers. Perhaps this could be structured more in a debate mode, where diverse answers could be grouped, and people can also answer to each other. I’d also like to be able to ask a question to the person who asked; for clarification. But I see that such forum style would make things much more complicated. Perhaps that was just my own perception. I liked the chats more.
Summary
Altogether, I must say that participating in I-am-a-scientist was a great experience. If you, as a researcher, have the chance to do that, I can only recommend it. It helps to think about the own work and the impact it has on other people.