Link to Medbot: m.me/103332218234386
Summary
This medical chatbot focuses on answering individuals’ questions about breast cancer. The sub-goal of this chatbot is to provide advice on issues the individual might have or can redirect the individual to a health care professional best suited to their symptoms. For instance, if the user has been experiencing symptoms of breast cancer they can use this chatbot to clarify their concerns. The main goal of this chatbot is to filter out individuals that believe they must go get urgent medical care at the hospital. Our purpose is to better direct breast cancer individuals to healthcare professionals or the hospital and at the same time spread awareness about health safety and taking necessary actions when required.
As mentioned above, the goal is to help reduce a significant number of individuals going to the hospital especially during the pandemic, hospital resources should be allotted to urgent care patients with Covid-19. This can also encourage time and energy towards doing research to finding the vaccine for Covid-19 because these days there is a significant increase in the people going to hospitals and it is advised by health care officials that it is better not to go to an emergency clinic/ hospital if the symptoms are not that serious and potentially reduce the risk of stress and panic created due to some false medical alarms.
Design Stages
During the design stages, we divided the design of the chatbot into stages, firstly we wanted to make sure that we have navigation with the Medical Chatbot as precise as possible. The design was made to clearly in fewer words start a conversation with the patient and speak to them regarding any symptoms they might feel regarding breast cancer, there was a several step design plan carried out for this prototype.
Firstly Research, we needed to pick the topic for our chatbot so before we went into actual research we brainstormed different possibilities (Figure 1). Cancer was the chosen topic, so we carried out research on different types of cancers and discovered that breast cancer was one of the most common types. We did further research on signs and symptoms most people with breast cancer suffer from so the bot could ask appropriate questions to the user. Due to the recent pandemic, hospitals, and emergency care have been occupied, thousands of patients admitted suffering from symptoms of Covid-19. Further research was conducted, figuring out which software would be used to implement the chatbot. We had to make sure that whichever software we chose could successfully connect to Messenger.
Figure 1 — Brainstorming topic for bot
Secondly, Define i.e. formulate the main vision and purpose of the chatbot and carry out a design process with the end goal of helping these patients with whatever they are feeling that particular day regarding breast cancer. We listed out user inputs that were found during initial research and then we added in inputs during team discussion and tied it all up with Chatbot and breast cancer (Figure 2).
Figure 2- Input on the chatbot
Third Formulate or Script, i.e. formulate ideas regarding the conversation this chatbot could have with the patient and more importantly the type of response you would get once you’ve interacted with her and shared your symptoms. According to Moorjani (2017), while designing a chatbot script there are seven things we need to take into the design stage i.e., Onboarding, ambiguous inputs, Verification, Resolution, switching intents Abandon flow, and Fallback. Next Onboarding, explained by Moorjani (2017), “Conversational UI can create additional cognitive load on users trying to figure out how they can interact with the bot, especially first-time users.” Examples of this from our bot would be: Hello! I am Medabot! I am here to help you!
Ambiguous inputs, we wanted our users to feel safe and for that, during our research, we found out that the more empathy the bot shows the better the chances are for the user to explain what they are feeling, we also narrowed down option selection for the only purpose of making the user answer.
Example: “How are you feeling today?” — “Are you concerned you might have breast cancer?”
Verification — Before we moved forward with the user and helped them diagnose themselves, we wanted to verify that the user was ready for the questions by specifically asking “Are you ready” and only continuing if the user said they were. We want to ensure that they are prepared for the questions they will be asked since breast cancer is such a heavy and life-changing topic.
Switching intents — We wanted the user to feel safe and help decide if they need to visit the doctor or not? The only way of doing that was including a sense of empathy in our script.
Example: “Are you experiencing irritation?” “Is there a rash?” “Do you feel a lump?” “Reply whenever you’re ready!”
Abandon Flow — After the user has got the response of if they should visit a doctor or not during our research we found out that this brings in stress in the mind of the user and it is highly possible they doubt the end result or want to go back and see all the symptoms, that is why we provided them with a link to learn more breast cancer. Figures 3 and 4 show the flow of the chatbot.
Figure 3- First version of the flowchart
Figure 4- Second version of the flowchart
And lastly Test (Figure 5), i.e. we had to do multiple test runs at every step starting from the initial greeting message to all the way to the end message to see if the chatbot was inclined towards the same prototype design we created. Our testing included all the team members and some family members testing out the Medbot with whatever they felt like choosing and then testing out the link to the Medbot and the link to the breast cancer website.
Figure 5 — Test Screenshots
Further Implications and the Process Details
Selecting the Chatbot Designing Tool
In the initial stage of our creation process, we used FlowXo but its features had limitations in meeting our expected design. For instance, we intended to generate final responses according to the amount of yes and no answers to the user inputted. However, FlowXo did not provide options to fulfill our design. We have also inputted levels of interactive responses in the FlowXo. Unfortunately, there was an error connecting with Facebook chat, so we had to choose an alternative chatbot creating tool, Talkbot.
Overall satisfaction on Talkbot was high. The tool offered a great deal of simplicity and convenience although the available features were much less than FlowXo.The conversational flow of our Medbot is constructed to provide concise and informative conversation regarding breast cancer symptoms, moreover, count the number of times that the user meets the criteria to provide solutions. Therefore, Talkbot was an ideal tool in creating custom messages (greetings, breast cancer information) and quick replies due to its simple and user-friendly design features and functions.
Testing Medbot’s Usability and UX
The optimal method to test the Medbot’s usability and user experience was by making our friends and family try the Medbot to understand its ability and limitations. Several issues arose regarding the questions’ quality, the validity of the contents, and lack of interactiveness. For instance, the tester one complained that certain crucial symptoms were not provided in the Medbot and the “conversation was too stiff”. Tester two also mentioned, “there is no room for the user to explain their experience” (lack of interactiveness) and the questions can be more creative in a user-friendly manner. Despite these flaws, the testers had satisfaction with the concept and level of informativeness of Medbot. The testers were also pleased to have the direct link to the breast cancer page in cancer.net (https://www.cancer.net/cancer-types/breast-cancer) and the aesthetic that Facebook messenger creates.
Figure 6 — Tester 1
Decision Makings and the Final Result
Our decision-making process was heavily influenced by generating the proposed atmosphere, feedback of the testers, and the available functions on Talkbot. Understanding from the perspective of a scared and anxious user was essential to identify the most suitable atmosphere to be created. The wordings and response order were changed to ensure Medbot delivers its responses in a friendly and caring manner to the anxious users. Moreover, we took the reviews of the testers into account, by adding more responses that checklists other crucial breast cancer symptoms. By incorporating the testers’ review, focusing on conversational usability, testing, and applying technical feasibility, Medbot lucidly demonstrates the vertical prototype. Medbot is connected with real data (testers), and it focuses on completing specific features in providing insightful and relatable information (Adams, n.d.). We understand that Medbot still needs multiple fixations, however, we worked towards achieving thorough details in completing the breast cancer diagnosis and ability to assess data (Users’ answers); thus, indicating the chatbot falls under the vertical prototype.
Strengths & Weaknesses, Possible Additions
One of the strengths of Medbot is that it is straight to the point while maintaining an informative nature to capture highlighting symptoms of breast cancer. Through these highly relatable and useful contents, the bot also analyzes the amount of yes and no answers to provide applicable end results and suggestions. Hence, Medbot provides quick scope on the essential breast cancer symptoms to those who are anxious, allows them to self examine, provides insights on possible outcomes, and suggests possible actions to take. Medbot also provides links to the website with highly informative breast cancer contents, in case the user wants to learn and examine more. It is evident that the examination process is much easier through Medbot since the users do not have to search and digest large amounts of content through search activities.
Medbot lacks in creating empathy and conversational flow with the users because it only provides structured questions with Yes or No options. We tried to compensate for its flaws by inputting as many dialogues with a friendly and caring tone so the user feels assisted and cared for. We believe that generating such an atmosphere is essential when dealing with heavy subject matters like breast cancer, yet lacking in interactive options and responses could potentially make the users feel less empathized, hence assisted.
In the middle of the Chatbot creation process, our group members came across an idea to add a geolocation feature so the Medbot can provide nearby hospitals or clinics that offer breast cancer examination and treatment, according to the user’s location. However, we recognized this function requires further knowledge and different designing tools that provide geolocation features (Google Maps), we were too further into the process to make alterations. We also wished to integrate a measuring system to assess the amount of Yes or No answers and generate the final responses accordingly. Unfortunately, the tools we explored had limitations, therefore we had to make the users manually count the number of the answers and choose the appropriate final results.
Conclusion
In conclusion, the Medical Chatbot aims to filter women that feel the need to go seek care at the hospital because they fear that they might have symptoms of breast cancer. Due to the pandemic, hospitals are flooded with urgent care patients, and PPE’s (personal protective equipment) is limited, hence the sub-goal of the Medical Chatbot is to reduce and filter out the less urgent or unnecessary hospital visits. In order to create this chatbot, there were many stages of designing before actually implementing the bot. First, research had to be on the topic itself, breast cancer. Then using the information gathered we could create the basic flow of the chat. Using a flow chart helped this process since it can keep track of what questions are asked and possible answers that could be received. Next would be choosing the software used to build the chat. As mentioned before we ran into several problems when using FlowXo so we had to experiment with the other websites provided to us. We settled on Talkbot and finally developing the chatbot could begin. Throughout this process, testing was a crucial component. Each change had to be tested in order to ensure the bot worked the way it should have. There were challenges with the creation but overall we were happy with the end result of Medbot, the breast cancer chatbot.
References
American Society of Clinical Oncology. (n.d.). Breast Cancer. Cancer.Net.
https://www.cancer.net/cancer-types/breast-cancer.
Adams, C. (n.d.). Interview Questions for Business Analysts and Systems Analysts. Retrieved October 19, 2020, from https://www.modernanalyst.com/Careers/InterviewQuestions/tabid/128/ID/5252/What-is-the-difference-between-horizontal-and-vertical-prototyping.aspx
Gibbsons, S. (2016, July 31). Design Thinking 101. Nielsen Norman Group.
https://www.nngroup.com/articles/design-thinking/.
Lucidchart. (n.d.). Intelligent diagramming for every team. https://www.lucidchart.com/pages/.
Moorjani, Y. (2017, April 27). Designing chatbots. Retrieved October 20, 2020, from https://uxdesign.cc/how-to-design-a-robust-chatbot-interaction-8bb6dfae34fb.
Dong Lee — 1000568496
Siddhaant Laul — 1005164983
Rachel leong Jia En — 1004604163
Shantal Fell — 1004360489
Nadine Smai — 1004448127