Embedded Files
Project Overview
In this project I worked alongside my classmate, Christos, to create a chatbot prototype targeting a specific use case for the MSOE website and present it to the Director of Digital Marketing at MSOE. We designed our chatbot to help new students get involved around campus. Our 3 main targets where events, activities and navigation.
While working on this I practiced my user research, conversation design, information architecture and interpersonal communication skills.
Process
Research: General Chatbots
To start, Christos and I conducted secondary research, looking into a wide range of chatbots. We first looked at chatbots like ChatGPT, Google, Siri and Alexa, trying to find what makes them successful.
We found that each of these chatbots is designed with a distinct personality, reflecting the brand identity behind them. Their personality traits are carefully selected and built to engage users, and more importantly, stay consistent across interactions. This consistency is crucial to build trust with users.
Each of these chatbots also has effective error handling. Our chatbot was going to need three features for effective error handling:
Clear Error Messages: When a request cannot be fulfilled or an error occurs, our chatbot should provide clear and concise error messages to inform the user of the issue.
Suggestions and Guidance: Our chatbot should offer suggestions or guidance to help users successfully navigate to their desired outcome.
Escalation Paths: When our chatbot can't resolve an issue, it should offer escalation paths such as connecting the user to a human or providing an alternative means of assistance.
Research: Benchmarking
With a better understanding of what makes the best general chatbots so successful we moved forward, deciding to look into other university chatbots to benchmark current solutions.
We looked into 2 different university chatbots (Fig. 1) & (Fig. 2). We were looking to find their interaction goals (what they were trying to help with), and their key personality traits. We also recorded a list of implications for our chatbots function as we interacted with them.
(Fig.1) Notes for Loyola Universities Chabot, LUie
(Fig.2) Notes for UC Berkely's Chabot, Berkeley
Research: User Interviews
Now that we knew what made chatbots successful, and had analyzed chatbots created with a similar purpose in mind, we wrote a script (Fig. 3) to interview students on our campus. We were looking to find what MSOE students wants and expectations from a MSOE chatbot. We talked with 5 students on campus.
The 3 biggest findings we pulled from these interviews were:
Students wanted the chatbot to be direct, and informative.
Students expected chatbot humor to be unfunny; don't try to be funny.
Students don't want to think about their interactions; guide the experience as much as possible.
(Fig. 3) Script written for MSOE chatbot research.
Define: Name & Interaction Goals
In an attempt to connect with our audience we named our chatbot R0SC03. This takes inspiration from the MSOE mascot, Roscoe Raider, Star Wars droids like C-3PO, and what is known as 'Leet speak' or '1337 5P34K' a language used by hackers in the 1980's and gamers in the 2000's.
MSOE is known for having a high population of introverts who often struggle with getting involved on campus. Between our classmates, friends and ourselves, we also knew that students have historically had a hard time with locating classes and events, especially in the Science Building. These are common themes in the qualitative data from this project and research from our other projects involving students.
We decided to aim R0SC03's capabilities at 'Campus Life' and be best at answering questions about Events, Activities and Clubs.
Define: Personality
Our qualitative data from the interviews helped us to define R0SC03's personality without assumption, and created a persona (Fig. 4) for our chatbot to organize it.
R0SC03 is:
Knowledgeable
Informative
Friendly
Direct
Helpful
(Fig. 4) R0SC03's persona, defining personality, tone, behavior, capabilities and some message examples.
Outline: User Journey
With the Name, Interaction Goals and Personality defined we were able to outline a User Journey (Fig. 5) for our chatbot. Brainstorming together, Christos and I broke down the journey touchpoints into 5 steps: Interest, Interaction, Engagement, Decision, and Action. This break down helped us understand our optimal conversation paths, and made sure our chatbot flows would be easy to iterate upon once they were built out.
(Fig. 5) User Journey created for our R0SC03 chatbot including: Touchpoints, Emotions, Critical Points, Solutions, and Functionalities.
Outline: Conversation Flow
Each step in our conversation flow (Fig. 6) with 3 possible user response options, following the 'Happy Path'. Within this initial conversation flow we explored potential error handling as well.
(Fig. 6) Initial conversation flows for R0SC03 chatbot.
Design: Initial Prototype
We built out our initial Voiceflow prototype, ideating further on our conversation flows. Feedback from our Professor guided us to lean into using of button response options. This was feasible because of the scope of our interaction goals and allowed our prototype would keep our users on the 'Happy Path' and reduce cognitive load for a majority of the interactions. To keep users engaged we made sure the chatbots responses did not exceed 27 words or 140 characters. We did allow the chatbot to chain separate response bubbles if it needed to address an error and prompting next steps. We were also able to implement 'intent detection' with the built in AI options from Voiceflow, feeding the back-end of our chatbot with response options that fit its tone so that AI generated responses would not diverge from the R0SC03 persona.
Research: User Testing
To work out any glaring flaws in our initial prototype we conducted user tests with 3 current MSOE students. We kept our script (Fig. 7) short, including 3 tasks. These tasks covered all of the chatbots functions and took less than 5 minutes. This was intentional to reflect a realistic experience using the chatbot.
The feedback from these user tests made us realize we were not providing the user with easy options to learn more about presented information. To solve this we added links to 'learn more' where they were applicable. This change as well as a few others are are explained further in the client presentation below.
(Fig. 7) The script/task prompts used to conduct user testing for R0SC03.
Deliverables
Client Presentation
My teammate and I presented this presentation to our stakeholders and our classmates. We created this presentation with the Director of MSOE Marketing as our target audience. The information was condensed for them, and the design was based on MSOE brand colors.
This was originally made in PowerPoint, but I uploaded it to my google drive to embed it here. For whatever reason, the text on some slides compresses even though it looks correct on google slides. This doesn't affect anything important, just a heads up.
Final Chatbot Presentation.pptxFinal Prototype
The prototype used to be embedded, but it would jump scroll to align the text entry whenever the page updated. Until I find time to solve this problem I recommend using this link to try the prototype instead! R0SC03 - Final Prototype - Voiceflow
Reflection
What I Learned
During this project I became familiar with Voiceflow and what it takes to design a functioning chatbot.
I also learned about the importance research plays on determining scope and target audience, how to keep users engaged within conversation design, and how to reduce cognitive load within conversation design. Working directly with our stakeholder also helped me understand how company guidelines and goals affect product development.
Project Completion Date: May 15th, 2023
Page updated
Google Sites
Report abuse