Montessori Compass Intelligent User Interface
Academic | Group Project | Spring 2019 | Course: HF760 - Intelligent User Interfaces
Assignment: Pretend your CEO has just asked you to add AI to your product. Choose an existing interface, and redesign it to make it more intelligent — in a way that brings real value to the product.
The Problem
Teachers in Montessori schools make detailed observational notes in order to track individual students’ progress as they follow children around the classroom to different learning workspaces. Schools are increasingly implementing digital records-keeping systems and requiring teachers to use them. How can we integrate AI into these digital records systems to improve teacher's experiences?
SKILLS HIGHLIGHTS:
-
Exploratory research/user interviews
-
Brainstorming/ideation facilitation
-
Wireframe design
Main Challenges
AI is too often implemented to ‘upgrade’ technology without thinking about the burden it can place on users.
The primary challenge was to find a solution that respects and balances user realities, the reality and nuances of AI, and stakeholder budgetary needs.
Research
Goals:
-
Understand the Montessori ‘way’ and what their processes are
-
Identify the most important parts of the process – the hows and whys
-
Discover pain points
-
Reconceptualize what the goal of the original digital tool was in order to help brainstorm alternative methods or solutions
Methods:
-
Literature review
-
User interviews
-
Group review of a commonly-used Montessori digital records system (Montessori Compass)
We conducted remote, in-depth user interviews with three current Montessori teachers.
As part of the interview process, we assigned our interviewees a 'homework' worksheet to bring to the interview to help them focus on the details of their days and processes.
We also conducted a general review of one current Montessori digital records system (Montessori Compass - the only one we had full access to) to gain an understanding of the type of system teachers are working with and how data is currently input and organized.
Key User Insights
-
Classrooms are hectic, with teachers observing about 10 students at once
-
Digital system has streamlined note organization and retrieval, but has created additional administrative burden for creating notes (have to retype handwritten notes into the system)
-
Teachers take notes on paper because it is the least intrusive method
-
Tablets in classroom are distracting to children, so not a viable solution
-
"Sometimes I stay [for 45 minutes after school is over] to just jot down some notes that I wrote on paper into the system.”
— User 2
Our User Persona documentation, highlighting key pain points
Proposed Solution
We created a fix for the main pain point teachers described - getting their notes into the digital system.
In our proposed system, teachers can input data into the system simply by taking a photo of their notes. AI algorithms parse the text into the system, and the user can review the data and submit it into the system with the click of a button.
Instead of manually retyping handwritten notes, teachers can simply scan their notes — saving them valuable time. AI algorithms will convert the image of handwritten notes into text in the input form, which the teachers can review and submit into the system with the click of a button.
Importantly, the proposed process does not change significantly for teachers, and the new notes-scanning step is likely familiar to other tech (e.g. scanning checks in mobile banking).
The Wizard Behind the Curtain: AI Technologies Used
step 1:
An Optical Character Recognition algorithm is applied to the photo of teacher's notes to convert handwriting to machine-readable data.
STEP 2:
Entity Recognition, Classification, and Text Extraction algorithms are applied in parallel to the machine-readable data, sorting the data into correct spaces in the review page (medium-fidelity wireframe shown here).
The Framework and Inputs for Training AI algorithms
The system will be trained with supervised machine learning techniques, with inputs such as student names, names of planned lessons and "works", etc.
The system should utilize continuous-feedback learning — as the user interacts with the system, mistakes may be corrected and algorithms re-trained in real-time.
(click image to enlarge)
Feasibility
We propose we can likely use open source tools for OCR, Entity Recognition and Text Extraction. This will limit the costs of implementation for the company. For Classification of works and progresses, a new algorithm can be trained relatively easily, as Montessori is a narrow field and the domain-specific language and categories used are universal across all schools.
Conclusion
Our solution:
-
Eliminates a pain point, saves teachers valuable time
-
Leverages teachers' expert knowledge; doesn't require end users to spend effort training/retraining the system
-
Aligns with long-standing behaviors and current mental models
-
Technologically feasible for small company; Good for school budgets (don't need to buy any new products)
PROJECT TEAM
Many thanks to my teammates on this project.
The team collaborated on many aspects of the project. Group members took leading roles in the following areas.
Katrina: Project leader, research lead (led collaborative development of research plan with first-year team members, moderated 2 of 3 user interviews), brainstorming facilitator, wireframe designer
Binita: Montessori expert, led initial group review/analysis of Montessori interface at kickoff, AI feasibility research, participant recruiter
Ryan: Big-thinker, interview moderator, report writer, presentation designer, interview note-taker