Tension Design: Chatbot for Vaccines

Initial Design

Our group started off with trying to find a tense topic. We brainstormed a list then had us pick the top three. We kept doing this until we ended up with one topic, vaccines. The group decided to work and create a chatbot individually and test it. Once we tested it we would go back and decide what is the best about each. Our goal was to have one that used buttons rather than parsed answers as it made the technical aspect of the project easier.

I worked on mine, in order to start I tried to figure out what the goal of it was. The goal of the chatbot was to be informative about vaccines, which means it is biased towards the usage vaccines. I wrote out topics I wanted the bot to cover, from history to how it doesn’t cause autism. I ended up with the flowchart below.

Flowchart 1

The next part was to work with FlowXO and get this idea to work. I wanted it to loop backwards, so if you were interested in only one topic one could read it and be done with the chatbot. But if you wanted to know everything you would still be able to do so. However there were some technical issues. This lead to flowchart 2.

Flowchart 2

It still had a similar premise, allowing you to see what you want. But it made it harder if you wanted to see more of the information.

Results of Initial Design

After the first round of prototypes we had three. Shreeja had an unfinished one, James had a mostly finished one and I had one. We looked and compared the ones James and I had. His resulted in a very neutral and sequential chatbot. Additionally, the bot had little personality. His bot would ask do you want to hear about so and so, then do you want to hear about this.

My chatbot had a very biased and aggressive tone. This might have been my own personal bias contributing to my writing. Mine also had a more iterative process to get the information in how you could repeat the entire conversation. However, while mine had a lot of content and arguments some of the grouping did not make sense.

Final Design

With those results, we ended up using mine as a basis but rewriting most of the conversational points. All of us worked on the rewriting trying to find a more neutral tone, but one that still had personality.

We also had to regroup some of the topics and added references to everything. While the ideas were ours, the content wasn’t. We made it less sequential and tried to cover most of the bases for an anti vaccine arguement. Flowchart 3 is the final flow we created.

Flowchart 3

With the topics rearranged and line to be changed, James and I worked on changing the actual FlowXO. He changed the conversational lines and added a name feature. This allowed a user to add a name and for the chatbot to use it in conversation. I edited the groupings and made modifications to the informational blocks. Below is a video of three different runs of the chatbot.

Results of Final Design

During the user testing in class we had three questions that we asked the user, what did you like, what is the personality, and what are improvements. I felt that having a chatbot with personality was very important as it makes the conversation seem normal and real.

In regards to what they liked they talked about information, conversation and looks. People liked having the bubble options to pick, but if they wanted they could still type the answers. If they typed wrong, the bot was able to catch that and give the question again. People liked how flowxo took time to type, creating a more human like persona. Several liked the amount of information that was conveyed and that it had references with more links. They felt that is was a good amount in total. In particular people liked the history information and the conspiracy topic.

Overall the users liked that the robot was conversation, that it asked for a name and used it. They liked how the bot asked if they learned anything and that the bot was easy to talk to. Some liked that the bot took a stance and carried through with that bias. Overall people liked how it was “well-implemented and fleshed out”.

Then the question of personality was answered. Most people thought it was nice and friendly. One pointed out that it was a bit sassy (they liked that) another attributed the personality to a friendly teacher. One did think that it was a bit condescending, with not much personality.

The main improvements they gave were about tone, options and large chunks of information. Some said that the bias could be jarring, but with the goal of wanting people to vaccinate, I feel that the bias is necessary. People thought that it would have been better to break up the big text into smaller chunks and feed in smaller pieces. This would humanize it more, phrase it like a conversation. Two did comment that they wanted more options, specifically in the anti vax arguments.

This was good for us as it made the changes from the first to the second seem right. That we changed things and made the bot significantly better. The improvements they did give, showed that we were on a good direction for the bot. We couldn’t have possibly covered everything about the arguments for and against vaccinations, but we made a bot that allowed people to easily use.

Link to chat bot: https://fxo.io/m/3z68nmpz