

Summit
Behavioral Design approach for Habit Formation
through Conversational Journaling and Habit Mindset
OVERVIEW
As part of my undergraduate thesis, I attempted to understand the science and psychology behind habit formation, and the role design can play to facilitate it.
I proposed and investigated two hypotheses - first, studying the effects of building habit mindset prior to habit formation; second, exploring the efficacy of a novel method of conversational journaling during the process of habit building. These hypotheses were realized in a final concept Summit, which takes users on a guided metaphorical journey of building habits. A functional chatbot was developed for testing the hypotheses through a between-subjects study with 30 participants.
ROLE
Design undergrad @IITG
Behavioral design, conversational design
research, conceptualization, UI design, prototyping, chatbot development, testing
GUIDANCE BY
Prof. Pankaj Upadhyay
Sustainability & Social Innovation Lab,
Indian Institute of Technology Guwahati, India
TIMELINE
18 weeks
Oct 2018 - Apr 2019
TOOLS
Miro, Landbot.io, Sketch, Diologflow, Glitch

Understand & Discover
Research Study comprising Literature Research, Market Study and User Research followed by their analysis
OUTCOME
Through the thesis, I conceived and investigated two hypotheses around Habit Mindset & Conversational Journaling. The outcomes spanned over two main areas - a functional chatbot to test the hypotheses via a between-subjects study; an app concept, Summit, to integrate those hypotheses in the habit formation process.
The app is based on the metaphor of climbing a summit where the user first trains (tiny habits), then plans the journey itinerary (habit plan), selects co-travellers & gears (habit partners & tools), & then embarks on the journey with a trek guide (chatbot that motivates, guides, & tracks progress).

Key screens from the conceptualized app

Sample conversation from the chatbot prototype
Design Process
The Spiral Design Process given by Stephen Wendel in his book, Designing for Behaviour Change, was adapted to guide the design process for the project:

Research Questions
As part of the research study, we originally set out to answer the following questions:
1. How are new habits formed and existing ones changed or stopped?
Understanding about science and psychology behind habit formation and behavior change; studying different models and theories proposed, and analyzing case studies and experiments based on them; gaining knowledge of the existing products & techniques.
2. How do users perceive habit formation and behavioral change?
Understanding the notion of habits in users’ minds, how they perceive and undertake habit interventions, the problems they face, and their attitude towards the same.
3. What should be designed?
Finding out possible potential opportunity areas in which habit intervention-based products can be designed.
Research Study
To answer the research questions, the following approach was chosen for the research ahead:

Literature Research
A very rich literature exists in the domain of Habits focussing majorly on the factors that constitute and influence Habits. An extensive literature study comprising key research papers, papers from conferences like persuasive technology, relevant books, articles, and a course on Tiny Habits by Behavioural Lab at Stanford was carried out. 5 overarching Habit Models were identified and studied along with their associated applications, case studies, and experiments.

Market Study
A set of 19 products that are currently available in the market were explored and analyzed. The techniques, frameworks, and methods used by these solutions were studied and the following themes were identified.

User Survey
The motivation for conducting a user survey was two-fold. First, to understand users and their behavior patterns; second, to derive opportunity areas by analyzing those emerging patterns and problems. With the need of studying these factors at a scale in a short time, we majorly relied on user surveys over user interviews.
A branched questionnaire about user habits, present intervention methods used, their shortcomings, and different factors related to habit models was framed. Some examples:
Why do you want to form this habit? (Motivation)
Do you feel like you can form this habit? (Perceived Behaviour Control)
Why do you feel you cannot form this habit? (Probing Questions)

Survey Analysis
Cross-users (Vertical) and Cross-questions (Horizontal) analysis was done to draw out insights.



Opportunity Areas
Based on the insights and gaps discovered from research study, different opportunity areas were identified. Five parameters were kept in mind: impact, alignment with habits discovered in the survey, novelty of the concept, scope of development and testing constraints (method, audience and duration).





Design
Conceptualizing and detailing of the solution including app and conversation design along with their prototyping
Problem Statement
After a discussion with thesis panel members and weighing the opportunity areas on the pre-defined parameters, the following problem statement was defined:
Conceptualize and design a solution to form and strengthen Keystone Habits and
Daily Lifestyle Habits.
Concept Directions
The conceptualization resulted in the formation of three potential concept directions. Each idea is based on a basic principle or technique upon which the whole concept is detailed out. Focus is laid on developing a novel idea or a newer technique other than the common approaches of setting goal, tracking and reminding.
In the ideation of the final concept, we tried mixing and matching different elements from the proposed concepts. We combined the conversational journaling aspect with gamification and habit-forming in groups. We also included the contextual triggers aspect but without the wearables. These elements were stitched together in a final concept using an interesting metaphor to add flavor.


Summit
Summit takes the user on a journey of developing new habits by building a habit mindset and retaining the same through a supporting self-narrative
Hypotheses
The concept is built around two hypotheses derived from our analysis of the research study. We would also be conducting experiments to validate these hypotheses in the later stage of the process.
1. Building Habit Mindset
Establishing habit loops help, but people tend to lose habits even after that. We propose that bringing people in the correct Habit Mindset before actually starting their habit formation can help us tackle this problem.
For achieving the habit mindset, we plan to use the concept of Tiny Habits for a period of 5 days along with providing some basic knowledge of how habits work. Gaining a basic understanding of how habits work, and achieving the initial small wins through tiny habits will help in establishing more motivation and self-belief to get a strong start.
2. Conversational Journaling
We propose the use of conversations in journaling and making people self-acknowledge their progress, evaluate and recover from mistakes, all by just asking right questions in the right way in the conversation.
We call it Conversational Journaling and propose that if done daily over a period of time, can significantly help in forming new habits.
Concept Detailing
A metaphor of climbing a Summit was designed to translate the ideas behind the final concept into the solution. The user embarks on the journey starting off with a 5-day training where he builds his habit mindset through Tiny Habits. He then details out the itinerary of the whole Summit and chooses/gets alloted his co-travellers to take on the journey together. A personalised virtual Trek Guide helps him throughout the journey with daily Campfire Talks. The solution was delivered in the form an app design and a functional chatbot.




As is clear from the concept, the solution can be divided into two parts - one, consisting of the overall app design which forms the base of the solution; second, the chatbot design controlling the character of Trek Guide concerned with daily conversational journaling and overall guidance throughout the journey.
App Design
Quick hi-fidelity prototype was designed. The objective was to translate and convey the concept and features through the basic app designs quickly, to save up time for designing and building the chatbot to conduct experiments validating the hypotheses.







Conversation Design
A chatbot was designed and subsequently developed for handling two aspects - first, the initial set-up and training period for building the habit mindset; and second, the daily campfire talk involving conversational journaling. Each of these would further help in validating our two hypotheses.
The basic process for designing the conversation flows was followed - starting from defining persona followed by conversation components, conversation modeling, framing script and dialogues, preliminary testing, refining, and prototype development.
Persona
The persona was designed based on the backdrop of trekking and habit formation keeping in mind the target audience of millennials.

Conversation Components
Conversation Components are the basic structural elements upon which the whole Conversation Model is built. The following conversation components were defined for the chatbot. These include the style and the manner of conversation, the prompts and intents upon which the whole flow and algorithm are based, and all the visual elements that go in the conversation from Chips to Rich Media.

Conversation Modelling
A master flow model was defined for the overall conversation. After this, smaller flows were built for handling each turn and possibility. All different possible scenarios were thought through including the edge cases. Errors and fallbacks for each turn were defined and integrated into the flow. Flowcharts following the same process were developed for four different conversation sets - introductory flow, setup flow, habit planning flow, and the campfire talk flow.

A preliminary test of the conversation flow was done through landbot.io prototype for introduction and training, and Wizard of Oz on Facebook Messenger for habit setup and campfire talk. After multiple iterations following models were finalized. There were four major categories of intents/inputs:
Normal Intent: Interprets what the user is saying and gets activated accordingly. Different sub-categories are defined accordingly.
Data Storage Intent: The intents to which the responses contain data that needs to be extracted and stored in the database at the backend.
Chips: These are buttons for quick replies for the user. These restrict the range of alternative flows, reduce the chances of error, and are quick to reply.
Fallback Intents: The error handling intents that get activated when a defined error occurs and bring back the conversation on track.


Refine
Analyzing & Measuring the solution’s efficacy
with hypotheses validation through experiments
Experiment Design
A between-subjects experiment was designed to evaluate the two proposed hypotheses around building Habit Mindset and Conversational Journaling. Group A participants would go through an initial 5-days training period of building Habit Mindset through Tiny Habits followed by a 7-days period of habit building of their choice facilitated through daily Conversational Journaling. Group B participants will directly start with the 7-days period without initial training. People who had previously tried and failed at habit-building would be approached for the experiment, to draw a comparison from their previous experiences to evaluate Conversatinal Journaling. The comparison between the two groups will help to identify the efficacy of building a habit mindset through tiny habits.

Chatbot Development
To conduct the experiment, a chatbot targetting the four conversation flows: introduction, set-up, habit plan, and conversational journaling, was developed based on the earlier designs. The introduction part was kept on Landbot. For the remaining three, conversations were handled through Dialogflow. Intents and algorithms were set up there based on the designs, which were further enriched with designed dialogues and their variations along with the rich media repository. Facebook Messenger was chosen as the chatbot environment which was linked to Dialogflow through a webhook developed on Glitch. The data from data storage intents is stored in Firebase.
The basic framework is explained below:


Feedback & Analysis
Insights were formulated through feedback forms and interviews conducted with test participants at the end of testing. The evaluation was done from the self-reported data by users in forms and interviews, due to lack of any other alternative mechanism for tracking habit automaticity objectively. Also, 2 participants in Group A and 3 in Group B dropped out midway leaving behind 8 and 17 participants respectively.

1. Chatbot Personality
All the participants liked the personality of the chatbot. They described it as very motivating, enthusiastic, happy, and cheerful.
“He is always so happy & cheerful. I have started looking forward to these chats now.”
They also found its portrayal as a chatbot instead of a real person helpful as they could escape their hesitation and feeling of being judged. However, there were a couple of instances where participants confronted that they lied and didn’t complete the conversation genuinely to cut it down short.
“When talking to a person in general, you always have some apprehensions in mind but knowing that it is a chatbot, you feel more comfortable in saying things as they are without any hesitation.”
People liked the overall tone, conversation style, and use of GIFs. However, there were four instances of technical errors reported. Towards the end of the period, some participants also found the conversations getting repetitive.
2. A feeling of someone being there for you
Many participants conveyed that while talking to the chatbot, they felt like someone is there for them who is trying to help them. This feeling of someone constantly keeping a check on them and pushing them in the right direction gave them the slight nudge that they needed. They even felt a feeling of not letting down towards the chatbot, showing ‘social norms’ coming into play.
“It feels as if someone is there for me who is concerned of me building the habit, I don’t feel like letting him down.”
3. Self-acknowledgement
Self-acknowledgement helped participants in many instances in getting back on track. In most of the instances when participants failed initially, they were able to overcome it within two days. Conversational Journaling helped them in acknowledging and finding right solutions to get back on track when they fail.
“It’s like I already know things, but it helps me get those things out or simply think about it.”
Also, when people started acknowledging their habits they started building self-expectations which gave them a sense of accomplishment and satisfaction when goals were completed and drove them to execute their goals.
“I started getting a feeling of completeness and satisfaction which kind of pushed me to do my habit.”
4. When the schedule gets tough, people crack
People easily get off track whenever some major task comes up, or schedule gets busy around the time of their habit. Although doing the habit would hardly take 5-10 minutes, they refrain from doing it. Although with time, when the habit got strengthened, the effect of this reduced.
“I had two back-to-back quizzes, I couldn’t find time to do it.”
5. Habit Building
The testing statistics help us formulate the insight that Conversational Journaling may be efficiently helping in building habits. But testing on a much larger scale and over a much longer duration is required to reach a solid conclusion.
The increased probability of building habits in second phase by Group A as compared to Group B clearly shows the efficiency of our first phase of building Habit Mindset in the overall process of Habit Formation. Again, testing with a larger audience and longer duration would be required to make a solid conclusion.
6. Increase in motivation and confidence
Whether they were successful or not, all the participants found the chats as helpful in reinforcing their behaviors and found themselves more confident and motivated for the next day after the daily chat.
Road Ahead
The preliminary testing as part of this thesis was limited to 30 participants and spread over a duration of 12 days due to logistical constraints. But now after the positive signs from preliminary testing, there is a scope for conducting the next round of testing with a much wider audience and spread over a longer period to statistically validate the two hypotheses. Prior to that, the four technical issues reported in the testing need to be fixed and more variations to the flows and dialogue need to be added to avoid the element of repetition over longer periods. If the validation with a wider audience also gives positive results, there exists an avenue for developing the end-to-end solution and launching in the market.
A research study around studying behavior nuances of millennials while interacting with chatbots is another interesting domain that can be explored. Building on this, the use of chatbots for learning and skill development can also be investigated.