Design for Tension
For this project of designing a chatbot on a controversial topic, we chose gun control for our issue. We set out to make a bot that would give a user the pros and cons of chosen topics from people for or against gun control. Real statistics from reputable studies were implemented into the bot to keep bias to a minimum. The point of the bot was not to persuade people on a side of gun control. Rather, it was to present facts and figures from both sides and allow the user to make their own decision of where they stand on the topic.
Brainstorming and Testing
Once we determined that gun control was the issue we wanted to tackle with this bot, we set out as a group to do research on the various pros and cons for and against gun control. Before we decided the bot should be neutral, we explored the pros and cons in depth to make a decision as a group on which side would be better for a bot to defend. We explored the option of creating texts from a bot that refuted arguments for and against gun control if the user entered them with different statistics. This was explored because we believed it would be better to make a persuasive bot than one that did not pick a side to the user. We initially thought it would give more personality to the bot rather than it just be something that gives concrete facts.
We tested these options by doing simulations where one person would pretend to be the bot and another would give it arguments. The person pretending to be the bot would refute it with different given facts. Eventually, we decided these arguments did not give the bot more of a personality. So after this we tried a simulation of a person asking the bot for facts on some arguments. We found this to be a much better solution and would help people in the middle of an issue better than the previous option.
Once we decided that our bot was going to be neutral, we laid out the pros and cons from eight issues and each user from our group set out to research one. We assigned the topics based on what each person wanted to research and gave a general outline how to make the flow so we could combine them at a later point.
The general outline was to define a list of triggers for the topic, have the bot ask whether the user wanted arguments for or against the topic, and gave the user facts. We would then link all of our topics together to create one bot that looped in order to keep giving information after one topic was chosen.
Combining and Refining
After everyone in the group completed their research and sections of the bot, we decided that some of the issues were too insignificant to be included in the bot. For example, the issue of a militia was cut because the arguments for it were weak and one side clearly had an advantage. Since we wanted the bot to be neutral, this was cut and refined to five selections for the user to choose.
Once everything was combined, we started testing the bot in our group to fix any bugs in writing, language and flow connections. After running through the bot a few times, we felt that the bot was too plain and formal with the user. We thought back to how successful bots like Siri feel like people are potentially talking to another person. To make the conversation seem more “real”, we added in some emotional reactions to the bot at some statistics like death. We also refined other topics that have no emotional language with transitions that helped the conversation flow better and seem more like a user is talking to a real person.
For some topics, we felt the need to condense the amount of information down into messages short enough to retain the user’s attention while still providing an ample amount of data. Therefore, we added in the option for the user to be given a link to external pages that expand on information on some topics we couldn’t fit in well to a texts.
The Final Prototype
When demoing the final prototype, users were able to find their way around the bot without any instructions given by us who were monitoring them. Every test user was able to get to a topic we had information on in the bot. We got feedback from users saying that they appreciated our approach to the bot being neutral instead of taking a side and that it was different than some of the other bots they tested.
Our final prototype chatbot on gun control successfully shows some key pros and cons of gun control and allows users to make their own decision on which side of the issue they want to be on. The strengths of the bot were being able to provide facts from both sides of the issue and have a good flow to it. Testers liked facts from both viewpoints on gun control presented and felt the bot was easy to navigate. The bot’s weakness was that some of the against arguments did not match the strength of the for ones or vice versa. If we had more time, we would do additional research to get more arguments for certain sides to make them equal. Despite this, users enjoyed the experience and we believe we presented enough information in our prototype bot for users to become informed on the key issues of gun control.