Discover Top Posts Tagged with #rnode

Arnod - A Chatbot for helping Raynaud’s patients

framing & strategizing; concept & ideation; creating & crafting; reflection & awareness; self-directed learning;

Introduction

In this project we designed a minimum viable product (MVP) chatbot that can potentially help patients keep track of their Raynaud’s phenomenon. The chatbot named Arnod, which sounds more humane - an important attribute of chatbots and conversational agents according to Radziwill and Benton, 2017, i.e. see Alexa, Siri. is an iteration on the smartphone application that accompanied my r-node smart ring concept of my bachelor thesis project (Partheniadis & Stavrakis, 2017). With the help of Lean UX strategy, Arnod explores the part of acquiring quantitative and qualitative data from the daily life of Raynaud’s patients to provide a usable, reliable, functional and emotional User Experience. A clickable prototype Arnod chatbot was tested with 19 Raynaud’s patients remotely using a new usability testing tool called Maze.design.

Initial Scope

This project started with the aim of redesigning the User Interface of r-node application which was the result of my bachelor thesis project. Primarily r-node aimed to provide Raynaud’s patients with an interface to (a ) keep track of Raynaud’s attacks and (b ) learn more about their Raynaud’s phenomenon. Secondarily, it aimed to (c ) bring patients together, (d ) bring patients in touch with doctors by sharing their data, (e ) bring researchers closer to patients to help them conduct studies related to Raynaud’s phenomenon.

Reframe of Scope

During the project, particular interest was found behind the idea of Conversational User Interfaces (CUI) for helping patients report Raynaud’s attacks. We identified high potential of CUIs and Chatbots to assist e-health applications (Chung & Park, 2018) in latest papers and applications (Your.MD, Sensely with Mayo Clinic, Buoy Health, Infermedica, Florence). From personal observations, more than 15,000 Raynaud’s patients are actively looking for support and solutions in self-organized Facebook Groups (1,2,3). The CUI could provide Raynaud’s patients with tracking and diary-keeping capabilities regarding their disorder in a familiar environment i.e. Facebook Messenger.

What is Raynaud’s

This personal project started with the aim of evaluating and iterating on the previous UX and UI that had been designed during my thesis project with title “Designing a smart ring and a smartphone application to help monitor, manage and live better with the effects of Raynaud’s Phenomenon”. People who suffer from Raynaud’s Phenomenon (RP) often experience cold fingers, toes and other extremities when they are exposed to cold or stressful situations. See pictures and clinical information here. It appears in 5% (Wigley, 2002) to 10% (Klippel, 1991) of the world’s population depending on geographic location, with women representing 90% of them (Nietern et al., 2015). The typical presentation of a RP cold ‘attack’ involves the fingers turning white (ischemia), then blue (cyanosis), and finally red (rubor). Raynaud’s Phenomenon is either described as Primary or Secondary. Primary RP occurs by itself (as a disease) by unidentifiable reasons and can’t be cured. Secondary RP is considered an expression of another underlying disease which, if identified and managed, it consequently cures RP as well. Both types have common symptoms, but the attacks can differ on symmetry, severity, frequency and duration and can be distinguished with clinical criteria.

Previous version of r-node app

The previous version of r-node features a lot of things like live temperature tracking(UI2,3), Raynaud’s attack prediction(UI2), attacks recording & history log(UI5), community of Raynaud’s patients(see third tab on UI2,3), activity tracking (UI3) similar to what other activity trackers do. This is complicated. This version has explored various different points of views around Raynaud’s. In this Iteration, its a primary goal to conduct validation research on this version by asking Raynaud’s patients, designers, engineers, e-health experts to give their feedback in think-aloud sessions. We are not aiming to evaluate this version strictly but we are seeking new knowledge, insights, challenges. Being in Netherlands gives the opportunity to connect with more patients and experts than I could in Greece.

Project Plan

March 8 - Validation Research: Subject A - Raynaud’s Patient: Meet a Raynaud’s patient from Netherlands, Run 3 Think aloud sessions with tasks, Validate current r-node solution, reveal strengths and weaknesses, potentially connect me with more RP patients.

March 9 - Validation Research: Subject B - Digital Designer with background in Arts: Aim was to see another designer empathize with the people who suffer from RP, see alternative point of view in terms of UX and visual language.

March 10 - Validation Research, Focus Group:Subject C,D,E,F - Digital Designers different disciplines: Aim to be generative and help

March 19 - Feedback from Irene Kamp: “Interesting to do a Lean UX iteration, Focus on what are the Insights generated for the patients and what is the System’s learning curve”.

March 21 - Validation Research, Expert Review: Pieter Kubbens, Neurosurgeon and e-Health Expert: Aim to connect with a network of e-health experts in Netherlands to (a) receive feedback on my project (b) potentially develop an application together.

March 25 - Validation Research, Reade Reumatologie Clinic: Aim to come in contact with health professionals to receive feedback from Raynaud’s specialists for r-node.

April 1 - Ideation: Visualize and Ideate on different point of views as derived from the validation research.

April 10 - Prototype: Filter out ideas and develop a prototype and prepare it for quick user testing using Maze.design.

April 15 - Evaluation: Publish for evaluation before may comes in. Post

April 30 - Evaluation: Gather results & analyze

Reflection & Portfolio Post

Desktop Findings

Top findings from the project include: Messaging is the number one way people are connecting today. The most important advancement in Conversational UI (CUI) has been Natural Language Processing (NLP) allowing the understanding of the actual human intent when they ask something from their device. More than 15,000 Raynaud’s patients are connected in Facebook groups to share their experiences and help each other regarding Raynaud’s. These Raynaud’s patients with an age range of 30-60 are actively seeking for a solution to keep track of they Raynaud’s attacks (thesis finding). Chatbots in smartphones make qualitative and quantitative assesment of one’s medical condition possible through a single user interface. According to Google Trends, chatbot is a trending search term and Facebook reveals chatbot-based conversations have increased 5.9 times since January 2017. Chatbots are being used more by people in their 40′s rather those in their 20′s. BI Intelligence reveals A Chatbot like Arnod (chatbot/digital assistant), can work together with the r-node smart ring (thesis outcome), to attain quantitative data real-time. Arnod can collect qualitative data in a strategic and natural way through its conversational user interface (CUI). Raynaud’s patients are familiar with Facebook and Messenger interfaces, since they gather in Facebook Groups to report and discuss about their Raynaud’s attacks, and we assume that they are likely to feel more comfortable using the Arnod conversational interface rather than a typical interface with text-fileds, sliders and drop-down menus. Trends in chatbot design in late 2017 provide users with buttons or drop-downs within the chat interface and allow text input only when needed, in order to minimize faulty inputs.

Validation Research Results

The low fidelity smartphone application confused most Subjects. The Raynaud’s tab should keep the amount of information shown to a minimal level. The subjects think the color-sensitive background space is very wisely used to visualize the risk of having a Raynaud’s attack real-time. Subject B noticed that the word ‘attack’ is not a good way to express the appearence of the symptoms as it automatically generates bad feelings (i.e. heart attack) and it might be important to work more on the branding and language used on the product application. Three subjects pinpoint the potential of the product to evolve as a research tool which can benefit the Raynaud’s patients in the long-term. They believe it has a lot of potential in helping study how human activities affect their Raynaud’s Phenomenon. Subject A, with RP, believes it could help identify patterns between people with common habits and common health problems and Raynaud’s symptoms and make it easier to share treatment methods they use. She believes an application like r-node can be a very powerful tool to reveal experiential solutions that people use to deal with their painful and embarassing symptoms. It has big potential to function as a place to track, learn and spread the knowledge about Raynaud’s.

Reflection

Strategy: The strategy worked out nicely. The Sketch below represents one Lean UX cycle from ‘Better Together’ agile framework. During the process multiple stakeholders(patient, e-health expert) and experts(designers, engineer) were interviewed, a chatbot-making tool was tested, and a remote evaluation was conducted with Maze.design and Google Forms with 19 participants who suffered from Raynaud’s Phenomenon.

Only one Raynaud’s patient was interested in participating in the test. I have been in close contact with a German male, two dutch ladies and a dutch representative of a Lupus association of Netherlands. Raynaud’s usually appears secondary symptom to Lupus patients. Although only one lady could finally come over. I prepared extensive tasks like holding cold drinks, warm drinks, typing on keyboard as a normal activity to observe how temperature normalizes. We went outside at a different context where the patient presented a short raynauds attack.

Validation Research: The project started with preparing the Validation Research of r-node. The research aim was to review the current product by patients and experts. The participant was 1 Raynaud’s patient, 1 E-health expert and Neurosurgeon Doctor from NFU, 5 Master Designers students and 1 Technology Expert. Contact was also made with Amsterdam Rheumatology and Immunology Center in order to arrange an expert review with a Rheumatologist who is more relevant to Raynaud’s Phenomenon. Unfortunately doctors weren’t interested in sharing their expertise, but they advised me to approach VU Medical Center, outpatient Rheumatology where they have a larger target audience of Raynaud’s patients. Due to lack of time VU was not contacted. For the validation sessions I prepared a sheet of 38 requirements on which the r-node would be evaluated. To conduct the test the following were used: a poster to explain the concept, working prototypes, cold and warm objects to interact, cameras. Data were collected as notes in a notepad or directly on the requirement’s list through a think-aloud process. The patient and one designer were left by themselves at the end to fill a typical System Usability Scale (SUS) form. The focus group consisted of multidisciplinary Designers. Subject C, 24 has a bachelor in Industrial Design, she considers herself quite social (3/5) when she feels comfortable with the people around her, has usually cold fingers but does not present Raynaud’s phenomenon. Her finger temperature during the test was 20 celsius degrees, and the room was 20. Subject D, 26, she has a bachelor degree in Media and a second one in Graphic Design. She considers herself likes to wear large rings and does not have an oversensitivty to cold. Subject E, 23, she has a bachelor in Design is very social person. Subject F, 25, he is a Product Designer and is a very social person.

Strengths

The user's and experts showed interest for the following things:

Sharing: Designers believe that the system should support the idea of community and sharing data. They suggest that the interface should support functions similar to what raynauds Facebook groups use like sharing images, discussing around them.

Eliminate Costs: Designers believe the patients will not willing to pay for the high cost (100-200€) of acquiring the r-node ring. They positively comment the opportunity to release a minimum viable product in the form of a smartphone application for helping patients keep track of Raynaud’s attacks.

Usership of the r-node ring: Designers propose a temporary ownership of the ring for the patients, since they are likely to get educated on how to treat themselves and then keep on living without the ring but with just the app. If the ring would have a starting price of 300€ she would be willing to wait and pay 100€ for a used ring if they are built to last.

Ideation

During ideation, the aim was to create quick sketches of how r-node would look like from the point of view of each evaluation participant. In total, 12 graphical user interfaces were sketched. The first two would visualize the commonalities between the point of view of the Technology Expert and the Neurosurgeon interviewed. Both have a strong technology background with internet of things, wearables and programming and are well aware of the GDPR laws. Their opinions strongly overlap about the requirement of proving the system with qualitative and quantitative data in order to better assess one’s state of Raynaud’s Phenomenon. Tech expert suggests chat-bots can help collect qualitative data from the user and, thus, help validate the sensor’s quantitative measurements. E-health expert underlines the importance of using questions as notifications in intervals which by his experience, helps collect useful qualitative data with little user effort. From my previous experience with computer vision and taking into account the trends of AI and machine learning, the system can translate qualitative data into meaning i.e. pictures into text for recognizing fingers and their color, or text into snippets to analyze and ask related follow up questions.

Quantitative & Qualitative Assessment

The concept behind r-node knowledge base is to assess one’s Raynaud’s Phenomenon by evaluating quantitative data with qualitative data. To put it simply, there are two point of views that can be combined in order to keep track of Raynaud’s events pragmatically and meaningfully: what the smartphone thinks I am experiencing right now? (sensors, numbers, predictions are quantitative data) and What am I actually experiencing right now? (human outlook is qualitative data).

Little evidence has been found regarding this process of evaluating mdeical conditions. Very recently, in March 2018, a paper points out that “Wearable devices for healthcare provide medical feedback to users based solely on specific biometric data” and thus “inaccurate information can be provided”. Subject H (Luis-Rodil) underlines the “paradox of automation” in which we confess technology becoming automated but the human factor becomes more important than ever to validate the system. The importance of qualitative assessment by humans is crucial especially for health applications in everyday life where no specialized medical devices are used to track the symptoms but regular smartphones. Quantitative data are collected by the smartphone sensors (location, activity) and the smart ring sensors (temperatures, humidity, heart rate, galvanic skin response). The system forms a hypothesis and creates qualitative questions to validate the hypothesis. The user replies or dismisses the questions which appear in intervals as notifications right after an attack has passed (according to the smart ring measurements) or after the user has entered a warmer environment or when the user starts to interact with their phone - which guarantees that their fingers are in a working condition. Qualitative answers are sent for analysis on machine learning and AI systems to help the system aggregate the information and present it in a human form. For example, the system may ask a simple but significant question to validate the accuracy of its measurements and help the system build a baseline to correlate which temperature equals to cold feelings for that person: “How cold do you feel your fingers in a scale from 1 to 5?”, where user simply types “4” in a numeric-only keyboard. A more difficult question could be “What do you think have caused this RP attack?” where user may start typing an extended report on what happened that moment or simply use voice input to say “no idea pal”. That’s part of data analysis and aggregation. After an extended period of time of use of the system we expect it to know more about which activities and contexts make the user more prone to having a Raynaud’s attack. With the combination of the wearable smart ring and the quantitative+qualitative assessment system, we expect data analytics and insights to appear about your situation in real-time: “Work gets colder 2 degrees after 14:00. Maybe it’s time for a tea!”. ”As part of their freedom of choice (GDPR), users can erase the quantitative and qualitative data at any time selectively or wholly. In the end, we expect the users to have gained knowledge from the contexts and activities that affect them or help them decrease the chances of an RP attack: food choices, exercise, places, etc.

Prototypes

Two prototypes were developed. The first was made in botsociety.io. Custom graphics where made in Figma and inserted in botsociety. Botsociety gave me the ability to quickly visualize the human-computer interaction with a chatbot called ‘Arnod’ - an easier alternative to r-node that also refers to a human name which are oftentimes used for smart assistants i,e, Alexa, Siri, etc.. With the first prototype I brainstormed on the user journey of ‘Reporting a Raynaud’s attack’. The aim of the project was to evaluate the system designed with users. Botsociety could playback the chatbot journey I had just created like a video which would make it easy to explain a concept to the stakeholders. Although, during user testing, we want the user to engage with the system and partake in the demo. Botsociety provides a small button that makes the prototype interactive! Unfortunately this did not function properly and as an addition I would still have troubles testing the scenario with users quickly.

For the second prototype, I chose to design the interfaces on Figma, based on Android Material Design GUI, with botsociety prototype and the best Facebook Chatbot Practices as a reference. The next step was to use InVision to make it interactive. For the chatbot creation it was important to use Timer in InVision; when a user says something, there is a delay before the bot behind the chat replies.

Evaluate

I used Maze.design to embed the ‘Arnod’ InVision prototype and conduct onlien user testing. Try the user test yourself, you might find it useful to try for one of your next projects. A link was created and shared strictly on Raynaud’s Facebook Groups. The Maze test had 19 participants in two days. Only 2 out of them completed the whole test. The rest abandoned earlier as the Maze data analytics show. This signifies three things: (a) firstly, users were unfamiliar such testing interface and needed more guidance and support by a physical person; (b) they expected the system to interact back when they would do a mistaken click and as a result they were easily annoyed and quit early; © they found the chatbot especially unfamiliar.

Overall, the users insisted on specific replies to respond to the Chatbot. The heatmaps that Maze generates are a little bit hard to interpret because you dont know actually the users intention. In detail, many times they wanted to express how they feel by themselves by clicking the text input. People where experimenting with all the answers although the heatmaps reveal that 18.2% of the users were avoiding to say “Hi!” to the Arnod Chatbot. The rest were feeling more comfortable about asking ‘what can you do for me’. The second question of the bot was ‘Do you have any cold attacks to report?’ and the people tapped either on ‘Yes’ or ‘no’ and noone tapped on ‘ ‘who are you?’ answer. The next question was ‘Have a picture of the fingers?’ and what’s interesting here to find is that several taps were recorded on the icon to go ‘Back’ or the ‘settings’ icon on the top bar and most taps were recorded on ‘Wait’ response! This could indicate that the users felt uncomfortable with the question as if they have to share their personal image with an unknown system. The question of the bot could be rephrased to ‘Could you please share a picture of the fingers?’ or ‘Would you like to register the event?’ and then ask for a picture where the intention of the bot is clear. When asked to select a picture some dummy pictures where shown. A lot of people clicked pictures that are not fingers which validates our need to use Google’s Vision API to identify that the picture uploaded is likely to be fingers or skin. Although people might still need to upload pictures of the context of the Raynaud’s attack e.g. car, so an additional question could be by the bot ‘I identified pictures that were not fingers or skin-related, would you like me to add them?’ *bot also shows these pictures*. When the picture of the fingers was uploaded, the question was ‘When did this happen to your fingers?’. A lot of clicks were recorded on ‘+’ as if they wanted to add more pictures which supports our previous suggestion to let them add more. Their second choice was ‘Earlier today’ or ‘Exactly at...’ to enter their own time of preference. People clicked that the event felt ‘So Painful :(’ or ‘Little Pain’ and less people clicked on ‘Embarrassing’. The system asked what triggered the attack and the automatic response was because of standing still and working too long. The automatic reply of the system was an insight and statistics, in this case that 62% get white fingers after sitting still and working for too long. The users firstly clicked ‘I don’t care’ meaning they really wouldnt care knowing about others in that moment. Secondly they would click ‘I can relate’ or ‘Tell me more’ and thirdly they would click on the bot’s insight message itself. So, its important to make insights clickable and link to proof of others’ similar events for more insights. At the end where the event was recorded the people wanted to ‘See more’ about the event and noone wanted to ‘Edit’ it. Maybe they could be one button as see more and edit there if a specific field is being clicked. I suggest trying the inVision clickable experience to see what happens on every screen.

After the users completed the task they were being congratulated, and they could optionally respond to a quick survey on Google Forms to leave their feedback. One user was very excited Maze is a great and a much requested tool by the community of digital designers. Conducting online quick user tests can save time and effort and help address the right target group through social media. It keeps track of time spent by the user interacting on each screen, counts and locates mistaken clicks with heat-maps. Maze requires more development in order to be recognized and trusted as a user testing software. It can give although big number of responses when you can send it to your target group.

Opportunity

Patient education for prevention is inseparable part for managing Raynaud’s successfully. It is considered that patients should take proper precautions for avoiding direct interactions with cold objects and environments. Multilayered clothing, glove liners, electric gloves, pocket heaters are essential for keeping the core body and extremities warm during the cold months. A healthy eating, non-smoking lifestyle with daily exercising or meditation can help improve circulation and relieve stress in the long-term.

Patients are poorly educated about what they can do for their disorder. According to all three RP patients during the validation research, doctors, usually Rheumatologists and Neurologists, do not provide enough resources to help patients learn about their disorder. Doctors give medicine only when the phenomenon becomes severe and they warn about side effects. Patients are used to rely on the doctor’s expertise and thus, they don’t seek for a solution themselves. The reversibility of the Raynaud’s effects make patients believe that it is not a chronic disease and because gives the illusion that it goes away. Patients get used to living with Raynaud’s and forget about how much it bothers them in their everyday activities. They learn to live with cold hands, painful purple fingers. They learn to live with it and when someone asks about it, they feel embarrassed. They can hardly talk about it. They avoid social interactions. R-node recognizes that people tend to suppress their feelings and wants to help them learn more about them. Upon arriving home after a cold day outside, r-node, through the user’s smartphone, may ask: “It was a cold day in Amsterdam today. How did you feel about it?”. The user can quickly start replying to the system or can select one of the predefined answers to the question that best describes their feelings: “So Painful :(”, “Little Pain”, “Embarassing..”, “Under Control”, etc.

Patients gather in Facebook groups with tens of thousands of members with different degrees of severity on Raynaud’s phenomenon. They exchange images, opinions, suggest activities to help reduce RP, share experiences and positive energy along the members. R-node recognizes the value of community, appreciates the simplicity of exchanging image posts, queries and links within Facebook and aims at integrating similar features in the smartphone application.With r-node we believe we can give patients a safe place to keep track of their phenomenon and also explore solutions and seek for help within the thousand other patients.

#mdd #feedback #personalproject #rnode #smart #ring #assistant #chatbot #botsociety #figma #invision