Your personal transit buddy at Carnegie Mellon University.
Client: Graduate Student Assembly (GSA) at Carnegie Mellon University (CMU)
Duration: 2 Months | Oct.2018 - Dec.2018
My Contribution: UX Research, UI/UX Design, Motion Design
Problem Background: Graduate Student Assembly (GSA) at Carnegie Mellon University (CMU) is expecting students to have a complete mental model of all transportation options available to them through the university, including basic information for use (e.g., location, routes, rules/requirements), facilitating the ability to contrast and compare options and efficiently make the best decisions for their needs especially for CMU transportation services (school shuttle and escort).
We received great feedback from stakeholders and users about how they can envision people or themselves could be using the help from ScotBot.
To understand students' needs, behavior and expectations when exploring and using the transportation services around campus, we conducted contextual inquiries to learn user's habits and mental model when inquiring for information, send survey to research on channels of information delivery and also think-aloud to discover breakdowns.
With contextual inquiries, we sought to understand how riders used and interacted with the shuttle and escort systems, identifying well-functioning flows as well as breakdowns. We conducted four contextual inquiry interviews across the primary transportation modes used by CMU students--walking, biking, PAT, escort and shuttles. In each instance, we accompanied our participants from the beginning of their journeys until they arrived at the last major junction by their homes.
We used affinity diagram (see it here) to categorize our interview data and to understand how information are spread throughout campus in different commuters' perspectives.
By distributing surveys we expected to learn about the demographics of different kinds of transportation riders, how they learn about transportation information and their priority when choosing the right transit method.
Learned about transportation information
through world of mouth (social network).
Did not go to all orientations whether
intentionally and inevitably.
Use Google Map and Transit as their main
tool of tracking traffic.
Chose PAT buses as their main
In order to understand students' behavior, pain points and breakdowns when they are looking for transportation information, we conducted Think-Aloud sessions with several CMU students.
Pain points we identified:
What we learned from generative research...
Based on our generative research findings, the main goal of the next phase is to generate potential solutions to address the current problems and also to validate those needs and how user would want the solution to look and feel like, which means what kind of service are they expecting.
Based on our generative research findings, we determined 12 potential needs from the users and brainstormed over 30 ideas according to the needs. After voting, we created storyboards that best address the needs:
Need 1: Identify optimal transit method
Need 2: Learn about all transportation methods (through an experiential way)
Need 3: Display and Compare of all available transportation options
To validate whether the needs we discovered were actual needs and also to evaluate the potential solutions, we conducted two rounds, 10 people Speed Dating sessions in total, validated most needs and gained new insights from the users:
Ideation Round 2
After we learned about their needs and expectation, we created more scenarios with solutions and conducted another round of Speed Dating to see which idea fits the best.
Among all the ideas that we generated to address this problem, one stands out as a friendly and smart solution, which is a chatbot who has all information and can optimize for the user.
To build a chatbot prototype, we started an account Facebook Messenger to simulate the process of using the chatbot. To begin with, we created slots, intents and prompts for the conversational system to build the decision tree. After finishing the prototype, we conducted Think-Aloud sessions to test the usability of the chatbot.
To build a conversational system, it is essential to analyze what are the:
See all the elements we built for the ChatBot here.
Based on what we explored from the last step, we built the decision tree for the conversational system:
In order to test on our prototype, we used Facebook Messenger to prototype our initial chatbot design.
Based on the feedback from our user testing, we iterated on:
The solution was appreciated by stakeholders who can foresee strong use cases of our product. For final iteration on my own, except for UI refinement I also envisioned a more futuristic design of the chatbot that featured with real-time tracking, machine-learning key word recognition, voice control etc.