MyST, My Science Tutor
Updated: Jun 11, 2023
My Roles: UX Research, UX, Conversational UI
Summary
"My Science Tutor (MyST) is an intelligent tutoring system designed to improve children’s excitement about and motivation to learn science, their ability to reason and talk about science, and their science achievement. MyST features conversational interaction with a lifelike computer character, the virtual tutor Marni, in rich multimedia environments. Conversations with Marni are designed to scaffold learning so that children can explain the science presented in illustrations, animations or interactive simulations.
- Wayne Ward & Ron Cole, Collaborative Research: Improving Science Learning In Inquiry Based Programs, National Science Foundation DRK-12 Grant 0733323), Boulder Language Technologies Technical Report, January, 2014, http://boulderlearning.com/wp-content/uploads/2016/12/NSF-MyST-Final-Report.pdf
My Part
MyST was an absolute privilege to work on, but I did sign a hefty NDA. Other than a brief summary of my work, I'll have to let quotations from the company and from a published NSF technical report speak for me.
My first role on the MyST project was as a student. I completed training courses for the Questioning the Author (QtA) reading comprehension format, as well as for the Full Open Science System (FOSS). Then it was on to the work. I designed and performed alpha and beta tests for the MyST software in a UX role. These tests took many forms. In part, I used an ethnographic approach where I tested the individual lesson plans alongside expert tutors and elementary teachers. We used the software in a group to gather user insights on what was working and what wasn't in a natural environment. This led to several insights into how changing the tempo of an individual session could improve the user experience. After the software was released to schools I studied hours of recorded children's sessions with MyST and devised multiple solutions to conversational and interaction design issues that were causing students to not finish or learn from their 15-20 minute sessions with the program. Some were obvious, such as the ability to mouse over an animation or picture and receive a tooltip hint about what the picture was. Others were less obvious... and proprietary. Eventually the conversational UI of the tutor was so compelling the children would attempt to engage it in conversation as a fellow human. Because of this I got to work on conversational tactics and decision trees that would bring the students back to the lesson without hurting their feelings.
Here's an early video of MyST in action.
A Longer Project Summary
The following is an excerpt from a technical report to the NSF, published as Boulder Learning was transitioning between grants during the middle of my tenure with the company.
"My Science Tutor (MyST) is an intelligent tutoring system designed to improve children’s excitement about and motivation to learn science, their ability to reason and talk about science, and their science achievement. MyST features conversational interaction with a lifelike computer character, the virtual tutor Marni, in rich multimedia environments. Conversations with Marni are designed to scaffold learning so that children can explain the science presented in illustrations, animations or interactive simulations. The specific objectives of the proposed MyST project were:
1. Develop, through iterative design-test-and refine cycles, a set of tutorial dialogs in which children converse with a virtual science in 4 different areas of elementary school science.
2. Create a corpus to support training and evaluation of the MyST system components— speech recognition, natural language understanding, dialog modeling, speech and language generation by the virtual tutor, and presentation of media within dialogs.
3. Conduct a summative evaluation of MyST to assess the feasibility of integrating the program into classroom science instruction, and its ability to improve students’ science understanding.
All of these project objectives were accomplished. Summative evaluations of two versions of the MyST system produced positive user experiences and significant learning outcomes, equivalent to human tutoring, and indicated the feasibility of integrating MyST into real world educational environments. Students were fully engaged in tutorial dialogs, and reported that they were more motivated to learn science after working with Marni. Teachers reported that they believed their students benefitted from MyST and that the tutorial dialogs aligned well with their learning goals." -ibid
Conclusion
The project is now over, I'm proud to have been a part of it. The results of all our hard work aren't just academic, however. I'll let Boulder Learning speak for themselves.
"The MyST (My Science Tutor) Children’s Speech Corpus consists of 393 hours of children’s speech. The speech was collected from 1,371 third, fourth and fifth grade students. The students engaged in spoken dialogs with a virtual science tutor in 8 areas of science. A total of 10,496 student sessions of 15 to 20 minutes produced a total of 228,874 utterances. 45% of the utterances have been transcribed at the word level. The MyST Children’s speech corpus contains approximately 10 times more English children’s speech data than all other English children’s speech corpora combined."
-
If you're a non-profit doing research, access to the MyST corpus is free.
I can see why building confidence through adult education would invite discussion, since different people will bring different experiences to the same question. A practical learning goal can make returning to a classroom feel more purposeful. I would keep that thought nearby when looking for ideas about adult skills courses. A short discussion of an everyday example would make this easier for me to connect with, especially as someone still learning about the subject. What would help an adult learner recognise progress early in a course?
I find planning an academic writing project easier to think about when there is a concrete question to start from. A clear outline can reveal gaps before a writer invests time in full paragraphs. That connection helps me see why academic writing planning can be worth a closer look. I would probably come back to this question after trying to work through one concrete example rather than deciding immediately. How do you decide whether a section supports the main question or merely adds background?
My first thought about developing confidence in language practice is that the everyday experience deserves as much attention as the larger explanation. A low-pressure conversation can help a learner use words they recognise but rarely say aloud. This is why I would approach language practice confidence with a specific question rather than a fixed expectation. I do not have a settled view yet, but this seems like a good place to slow down and look more closely. Which activities encourage speaking without making every mistake feel important?
When it comes to supporting teachers through shared learning resources, I would rather understand one useful detail properly than rush through a long list of ideas. A resource becomes more useful when it includes enough context for another teacher to adapt it. That is the kind of detail I would keep in mind while exploring teacher learning resources. I am more curious about the process than a polished result, because the steps in between are usually where questions appear. What information would you include when sharing an activity that worked in your classroom?
There is something approachable about discussing turning curiosity into deeper understanding through situations people can actually picture. A good follow-up question can turn a quick fact into something worth exploring. It also gives the phrase curiosity and knowledge a clearer meaning than a label on its own. I am still thinking through the details, and a small example would help me picture how this works in practice. How do you decide which question to investigate after an initial explanation?