Talking about Euthanasia: Educating College Students through a Chat Bot
For this design sprint, our team was asked to design a chat bot to deal with a topic of tension. We ended up designing a chat bot named Crystal who tries to educate people about Euthanasia and ask them their opinion on various cases/situations. The demo video below gives an example a user interaction.
Our goal with the chat bot is to educate college students about the various types of Euthanasia and how they differ from each other, in a more interactive manner.
While looking up ideas for this project, we wanted our topic to be something that was a little bit uncomfortable and invoke a reaction from the user. We came up with some initial ideas out of which, we narrowed it down to three: NFL players kneeling during the national anthem, Disciplining your children using physical violence, and Euthanasia. We ruled out the NFL situation because we thought that it would not be conducive to a meaningful conversation.
We thought hard about child discipline since when we brought it up we all had different experiences. However, when we did some research on the topic, we found that all of the information basically pointed to the idea that any physical punishment of your child is detrimental to their development. After we researched euthanasia, we discovered information about the categorizations of euthanasia which were interesting, and thus decided to proceed with it.
We use the 5-step design process found in the article “Designing Chatbots” by Yogesh Moorjani, the link to which can be found here.
1. User Intents — Scope
Since none of us really knew much about euthanasia to start with we thought that our age would be a good demographic to build our chat bot for. Instead of trying to have a discussion with the user and show them the merits or faults of either side, we decided that our chat bot would inform them about euthanasia and talk to them about their personal feelings towards the matter. This was a crucial step in our design process because before this we were feeling overwhelmed with the possibilities of conversations that our chatbot may need to deal with. This scope helped us reduce the number of flow decisions without negatively impacting the user experience.
2. Key User Inputs
One of our key user inputs was whether the user had any ideas on Euthanasia or not. In addition, we wanted the user to tell their stance on euthanasia (whether they support or oppose it) and explain their stance.
3. Play Assistant
We all went out and had some conversations with our friends about euthanasia to get an understanding of what the actual flow of a conversation about euthanasia would be like. We had general conversations with our friends where we talked about euthanasia. In addition, we also chatted with them (via Facebook messenger) while acting as chat bot assistants. One of these interactions is given below.
At this stage, we also determined the persona of our chat bot. We decided the following characteristics for our chat bot:
- Name: Crystal
- Gender: Female
- Age: Mid 20s
We thought that users would respond better to a chat bot with a female persona and with an approximate age range that is equal to the age of our intended users to be able to connect with them better.
4. Design Flow & The Script
We decided to combine steps 4 & 5 since we already had an idea of how to frame our questions from our experience while Playing Assistant (Step 3). Coming back together and combining what we learned from those conversations we developed the start of our flow. We developed only the initial half of our flow to start the conversation, and determine how much the user knew of euthanasia. After establishing that, we asked them how they felt about euthanasia and why they felt that way.
Once we established this flow, we then used Flowxo to create a chatbot to implement it. This was our chance to start working with Flowxo and get acquainted with the technology. There was a learning curve with the software since it presented most of its features in a linear manner. The linear orientation was difficult to work with since, as you can see with our flow, we had a two dimensional graph to depict it. We were able to implement all the features that we wanted to with our flow using Flowxo. Once we saw that we had a partially working chatbot, we were ready to test it out. We wanted to to see the direction users took the conversation after “Thanks for sharing”. Once we saw where users took the conversation, we wanted to implement the flow. We decided that it was probably better to wait and test it out than to already create a flow and take the user towards a conversation that wouldn’t be meaningful.
After testing it out, we saw that it would be more beneficial to provide hypothetical situations and ask the user again for their opinion on the topic. Initially we wanted to move the conversation towards religion, but we decided that it wasn’t engaging the user enough. It wouldn’t challenge their thoughts and make them think about why they support their opinion. Thus we would initially ask for their opinion and reason for their opinion. After, we would provide hypothetical situations to engage the user’s reasoning.
We see that this is an example of putting the user into a position to use their reasoning to justify their opinion. Our bot launches these hypothetical situations towards the user to see how they justify their opinion. Through experiments from having natural conversations first, we realized that the common trend in our person to person conversations are to bring hypothetical situations up and see how each other reacts to that. This is what we wanted our bot to accomplish.
We determined that we were ready to move onto the second half of our flow where we would get more in depth in our conversation about euthanasia and challenge our users to justify their opinion. Doing research into euthanasia, we found that there were actually three different classifications:
- Voluntary Euthanasia, when the decision for euthanasia is made with the patient’s consent.
- Non-voluntary Euthanasia, when the patient is unable to give their consent.
- In-voluntary Euthanasia, when euthanasia is performed against the will of the patient.
We thought about the best way to talk about these three types and the idea we came up with was using scenarios. We found an article online that depicted each of the three cases of euthanasia. We wanted to present them in our chat bot as hypothetical situations where we could ask the user what they would do in each situation. After asking a couple of questions to them about the situation and their answer, we had our chat bot send them a link to the article so that the user would be able to read what actually happened in real life.
We see that in the chats, the bot was pushing the user back and forth to justify their opinion. This seemed to us to mimic a natural conversation like the one we had in person with other people to determine our flow initially. Once we saw this going in a direction that we liked, we implemented the final part of the chart to illustrate the rest of the flow.
After developing the second half of our flow, we implemented it in FlowxO. The second half of the implementation was easier due to the experience that we gained from implementing the first half. Once we had a working model, we started our user testing for the entire chat bot. For this user testing, we decided that we would use people who we were not as close to such as acquaintances from class instead of close friends. We wanted our user testing to be removed from close friendships so that we could get some accurate representation of how almost a complete stranger would interact with our bot.
Feedback and Future Work
The feedback that we got were mainly small issues the user’s answers triggering the wrong items. This was easily fixed by working in flow and addressing the issue. The other feedback that we got was that during the presentation of the scenarios, the wall of text was very long. We worked on slimming it down and also breaking it down into multiple messages so that it seems more like a normal text conversation.
In the future, if we continued with this bot, we would like to make the conversation more two sided. We would want the user to probe our bot with hypothetical situations and put our bot into a position to justify her view. Our conversations seemed long because Crystal was the one driving the conversation, if our bot was more functional and responsive, conversations would be more natural. Thus allowing Crystal to handle more diverse responses from the user will definitely be an improvement.
Our chat bot was able to address the issue of euthanasia and present various scenarios to the user to educate them further. Through the various user testings and demo version in class, we are pretty sure people got to learn something new about a topic that normally people would have been uncomfortable addressing. Though the scope of the chat bot was limited due to time and technology constraints, we believe our chat bot has potential to expand and cover even more topics within euthanasia given more time and direction.