How well do you know about Women’s Health? Try our Bot!
In an age of information, it can be difficult to distinguish fact from myth. Women’s health in particular suffers from rampant misinformation due to a lack of education in schools and the stigma surrounding discussing women’s bodies. Our team (Sofia Gukelberger, Amanda Chu, Yijun Liu, and Sichang Tu) created a Women’s Health Bot via Flow XO that educates on different aspects of women’s health issues by asking the user to take a 10-question quiz, therefore clarifying any misconceptions and prompting the user to evaluate their own assumptions about Women’s Health.
1. Background and Motivations
1.1 Why Women’s Health?
We have come up with several ideas for tense topics (including mask mandate, vaccine hesitation, gun control, the current Ukraine war, and others) we finally decided to tackle women’s health for the following reasons:
- As an all-girl team, we found this topic quite self-relating. For example, Sofia had personal experience in seeing misinformation about women’s health being spread on social media. I liked this topic in particular because, as a Sociology major with a strong interest in gender studies, many of my courses have introduced how women’s health and bodies are misunderstood due to cultural and historical reasons. Additionally, I have heard from friends (both women and men) having misconceptions about women’s health, particularly about sex/STD-related topics. Therefore, I wanted to use this bot as an opportunity to learn more about spreading correct information about women’s health.
- Women occupy half of the human population, yet their health problems are often misunderstood. Thus, we believe that women’s health is an important issue to address considering the amount of population that is affected.
- Women’s health is very factual and thus we are able to provide an accurate and solid answer to the questions and concerns. Therefore, we eliminate the human biases within our team in developing the script and leading our uses due to our personal experiences.
- Due to technical constraints, we could not import a large database nor perform machine learning. Thus, the conversation would be less natural and can easily end if the user did not follow a typical script. Therefore, we believe that, given the technology, we are using, building a bot with an educational purpose/possible a quiz format would be promising.
1.2 Why does there exist misinformation about women’s health?
To understand our topic, we first need to know why there existed such problems and even political polarization in women’s health. After preliminary research, we have suggested several aspects:
- Not enough education for men. Men don’t think they need to be concerned about women’s health (don’t view women’s issues as human issues) and thus spread false information sometimes
- Women sometimes were told false information because the patriarchal society politicalized women’s bodies and choices, ignoring their autonomies and objectifying them. Additionally, men who are affected by patriarchy and heteronormative are ignorant and are unwilling to learn about women’s health. For example, sex is viewed as something that alters a woman’s worth, which is linked to the worth of her body; and there is not enough recognition of women’s pleasures in sex
- Echo chambers — eg. if you believe in something like on the scale of how much autonomy women have over their bodies, there’s a tendency to follow the same information. Likewise, users tend to congregate around similar users with the same beliefs. Therefore, it is hard for users who have false information about women’s health to gain the correct information.
Therefore, it seems that the problem is not only a lack of education but from a variety of aspects including cultures, genders, beliefs, and friend circles.
1.3 Who are we targeting?
Since women’s health is an important and fundamental topic that everyone should acknowledge, we tried to include every group that has misinformation about women’s health due to a variety of reasons. According to our findings above, we identified the following groups:
- Groups that engage with sexist internet spaces
- Groups that are self-relating (women) and curious about whether their information is correct or not
- Groups with strong beliefs/convictions about policies related to women’s health or autonomy over women’s bodies
- Groups who don’t usually look into women’s health (eg. populations that didn’t have that educational background, men, etc.)
- Groups averse to fact-checking
In short, we aim to target groups that may 1. ignore factual information about women's health (due to gender, culture, friend circles, or personalities)or 2. do not have much access to women’s health facts.
Thus, their expectations might be 1. feeling that they know everything about the groups who ignore factual information about women’s health 2. wanting to take the bot as an educational resource to learn more about the group who do not have much access.
Additionally, to achieve this broad audience, we decided to implement a web-based bot that can be shared easily across platforms.
How might we build a Women’s Health Bot to help ignorant people and/or people with low health information access to know more about women’s health?
The solution to informational-wise problems is always simple and straightforward: education. However, the word “education” is too broad and we need to specify it. Therefore, we first came up with a list of our goals here:
- Broadening information exposure by presenting a wide array of facts
- Trigger people’s desires to learn about women’s health
- Be conversational and friendly to avoid driving the user away
To tackle this and try to engage the targeted groups and achieve our goals, we proposed that we should design a bot where the primary function is to have a quiz to test users’ knowledge about women’s health. There will also be an overall score and a sub-section score (for different aspects of women’s health problems) for users to share with.
We believe that the quiz format will work better because by taking a quiz and seeing a result, we are not trying to directly educate them but to inspire
We hope that the quiz format can help:
- For those who ignore women’s health problems: hopefully, they will find it interesting to have a score/share with their friends about their knowledge of women’s health after seeing others sharing it on social media; by receiving a low score, we hope that this group may recognize their deficiency in women’s health knowledge
- For those who do not have much access to women’s health: hopefully, this quiz can serve as an assessment tool to tell the group how overall they know about women’s health and what specific categories they are lacking
As mentioned in the background section, we proposed to create a responsive quiz to help the two major groups we identified above. However, we still need to decide on the details of our quiz. We firstly did research on our own and each of us should come up with 10 questions with references. In the meantime, we did some pre-bot user interviews and testings.
2.1 User Research
2.1.1 User Interview
Firstly, Amanda did a user interview before we start building the bot, and she found:
- Currently, the typical solution to information problems in women’s health is googling information and not trusting posts shared on social media blindly
- The bot should focus on facts and exploring misconceptions
- Should provide links and resources that are trustworthy
- Bot should be friendly in misconceptions and not make the user feel bad if they got it wrong (some users don’t INTENTIONALLY seek bad spaces but false info spreads easily)
Together, with our discussion on the bot, we decided that, when the bot approve/disapprove the users’ answer, it would also provide a short explanation and relevant links to trustworthy articles to increase credibility. Additionally, we decided to provide users with a score that can reflect their overall level of understanding of women’s health. The score will be x/10, where users will earn a point when they answer each question correctly.
2.1.2 Number of Questions, surveys
I surveyed 12 people on how many questions they wanted to answer for a quiz about women’s health without quitting in between, and the data is the following:
5 questions: 2 people
8 questions: 4 people
10 questions: 4 people
12 questions: 2 people
As most people prefer 10 questions, and there is not a huge difference in the number of questions between 8 and 10, we decided to use 10 questions for the quiz.
Nevertheless, we still have disagreements in terms of the kinds of feedback we should provide (such as when to display right/wrong, immediately, or only an overall score) and whether there should be a quit button displayed each time. Therefore, we decided to tan some user tests afterward, specifically via the Wizard of Oz method.
2.2 Bot Persona
According to the interview and our discussion, we tried to build a bot who is very friendly and helpful. It uses clear, conversational language, yet professional and formal in explanation at the same time. This way, the bot would seem to be professional and reliable, and people thus would believe the information from the bot. At the same time, the bot will not annoy people who got wrong answers and quit the bot. We try to build a relative neutral bot in regards to people’s knowledge in women’s health. Additionally, the bot should not be a “dry scholar”. Therefore, we try to include several emojis to be fun.
We tried to change the wordings to avoid jargon, explain every possible terms that people may not understand, and try to be short and simple so that people would actually read the explanation.
2.3 Ideation Catch-up
To quickly summarize the findings above, at this stage, we decided to build a 10 question, quiz-based bot about women’s health. At the end of the quiz, they will be able to see their total score and receive comments about their total score from the bot. Additionally, we also want to provide a page of relevant links to users who do not want to take the quiz.
2.4 Formalizing Conversation
Since our bot is a quiz-based bot, there are not many keywords to match or important orders in conversations. Instead, we need to formalize our logic in building the bot, such as how to quit, how to navigate the quiz, and the information (link) page. Therefore, we have come up initially with the following flow charts:
The flow chart we have come up with was quite similar. We all have a start menu and a quiz section. During the quiz, if the user wants to continue the quiz, the circle will repeat until the quiz questions are exhausted. The major difference is that Sichang’s flow did not show details about the quiz, I didn’t consider other options such as resources and separated the quiz into sub-categories, and Amanda did not have other options as well. After discussion, we decided to adopt Sofia’s flow for formative user testing.
3. Formative User Testing
3. 1 Initial User Testings — Wizard of Oz
3.1.1 Flow of the Bot, Wizard of Oz
Amanda did two Wizard of Oz tests via Facebook to test how we should present true/false users’ answers.
- For the first participant, Amanda did 5 questions and gave them a grade after; and for the second, Amanda did 8 questions and told them right after if they were right or not.
- The first person said getting the grade after made her forget about the initial questions and had to scroll up. Would’ve preferred knowing if she was right or wrong right after answering
- The second person said the links to reference is interesting because they weren’t sure what would prompt them (if they were the target user group) to click on the link
Therefore, based on the results, we decided to send the answers and explanations right after each question.
3.1.2 Quiz Button, Wizard of Oz
Sofia split users into 2 groups, where Group 1 would be asked after every question if they wanted to quit and Group 2 would be told at the beginning that they could quit if they wanted to at any time.
Group 1 answered ~2 questions
Group 2 answered all questions
Completed bot user test results:
Liked that it gave resources
Liked the emojis
Was very engaged in the quiz
Sometimes it crashes on the bot info page (doesn’t like multiple people viewing the site at once?)
Therefore, based on the results, we decided to not include a quit button each time, but tell users at the beginning that they can quit by typing the word “quit” to the bot.
3.2 General Feedback — Design Goals
During the Wizard of Oz testing, we also received several general feedback:
- [+] Liked that it gave resources
- [+] Liked the emoji’s
- [+] Was very engaged in the quiz
- [+] Flow naturally
- [+] “It could be impactful for women raised in an environment where these sorts of topics are taboo, allowing myths to be perpetuated”
- [+] Like the scores to have a general idea of how I did it
- [ — ] provide links but not hyperlinks that can be click on
- [ — ] Want to know in details about what I didn’t familairze with
At the initial stage, the bot only gave general feedback and did not have a sub-section score. A common comment I received is that the bot is too easy and they want more detailed feedback. Therefore, we decided to add a sub-section score and thus giving the users a better understanding of what type of information they were missing out.
Therefore, we believed that our design has achieved our goals of broadening information exposure, triggering people to learn about women’s health, and being conversational and friendly. Yet, we still have aspects to improve. We decided to include a sub-section score, where we categorized our questions and would display to users if their sub-section score was lower than 50%.
4. The Bot: “Women’s Health Bot”
4.1 Logics, Flow Chats, and Conversations
Firstly, we formalized our flow chart and developed the complete logic for the bot so that users can always go back to the main menu and find where everything is.
According to the formative testing, we have also formalized the flow of the quiz’s part:
We have formalized our script into a friendly yet professional tone as well:
4.2 Final User Testing
4.2.1 How to display information?
After I finished a rough draft of the quiz bot, I tested 5 people on the ways to present information to decide which to include. I am also observing their behaviors and asking general questions about what they like/dislike about the bot.
4/5 prefer the second one (combined post) over the first (separated) one. The main reasons included: when the bot displays multiple posts, they came out simultaneously and are difficult to see when messages are poping very quickly. Additionally, it was harder to find information and need to identify what each part means. However, the ones that preferred the separated messages believed that it would be easier for them to identify their current score.
4.2.2 General Comments
We have also received several general comments after we decided to use the combined posts:
- [+] They said they liked that the bot was smart enough to prompt if you forgot to answer.
- [+] Liked that could type or click on the answer
- [+] Overall liked it and did learn from false misconceptions
- [+] Felt like know too less about women’s health and probably need to know more
- [+] Correct my misinformation in the period that I read on the internet
- [ — ] Questions felt a bit specific
- [ — ] provide links but not hyperlinks that can be clicked on
The above user testing provided very positive results. However, due to technological constraints, we are unable to implement hyperlinks in Flow XO. We tried to include the links using cards, however, since we have multiple links for each question, we need to display them all. Yet, Flow XO only allows us to display one card at a time.
4.3 Demo Video & Samples
Finally, we created a demo video to capture the important conversational features of the bot.
We also try to provide bot info about the bot creators and the class information.
As mentioned previously on the ideas of separating into different categories for quiz questions, we also separated the health recourses for the links because there were a lot of links.
The bot will display information about the category that the user selected. It will firstly randomly display an article (links that users can click with a description) from all the articles mentioned below, and suggested the users start with this one. We believe that it would be more user-friendly than just the plain text in links.
However, due to Flow XO’s constraints, we can only display one article in the format above (with a picture). Thus, we can only display plain texts for the result of the links, including the links, website name, website type, name of the article, and a short description.
As mentioned before, the quiz section will begin by telling the users that they can quit anytime by typing a quit, and will display all the sentences (correct, explanation, score, and read more) in a combined format.
At the end of the quiz, users will get an overall score, a comment on their overall score, and if there are any, a comment on the potential categories that the users are not familiar with (with a 50% less correct rate).
If the users want to know more (by clicking yes), the same information from the health resources section will be displayed.
5. Reflection & Conclusion
Overall, I believe that our bot is very successful in achieving our goals of broadening information exposure, triggering people to learn about women’s health, and being conversational and friendly. We completed comprehensive user testing throughout and constantly received positive feedback and improved upon the negative feedback.
However, I still believe that our bot has its own limitations. Firstly, although we did user testings throughout, our users were not mainly our potential user groups. My classmates and friends are informed about women’s health, even my male friends. They usually got pretty high scores (7/8/9), and thus the sub-section comments usually did not appear. They found the bot interesting, yet we wanted to know more about our targeted user group and their experiences. I believe that we need to improve on this aspect in trying to find users from our targeted user group.
Additionally, I found difficulties in using Flow XO as a free member. We could only build 3 flows, and in order to have the function to “return back to the menu”, we have to repeat building the menu flow three times. I found it very time-consuming and made the bot’s logic confusing as well.
In short, I enjoyed the design process for this conversational interface. It is my first time designing a bot, and I learned everything from the scratch. I understood more about logic, conversations, and user testings.