Teaching kids how to write by using a tablet

This app’s goal is to enable children in developing
countries to teach themselves basic reading, writing and
arithmetic within 18 months.

My Role

Global Learning XPRIZE is a $15 million global competition initiative to provide free education to all. The goal of the contest is to come up with an Android tablet application that will teach kids how to read, write and do basic arithmetic, specially in areas where there is a lack of instructors.

I worked with a team of developers and psychologists at Carnegie Mellon to come up with the finger-writing interface for our app, Robotutor. This interface will teach students how to write digits (letters and numbers), words and sentences.

Project type:
Educational technology
6 months (part time)
Team members:
Prof. Jack Mostow (team lead), Vishnu Rajan Tejus, Kevin Willows, Iris Wu
Tools used:
Illustrator, Photoshop

The Challenge

Besides having a challenging problem to tackle, there were a few aspects the team had to think about. We knew the Robotutor’s students would have no formal education, therefore making it difficult to assess their previous knowledge. Because of this, we had to assume they rarely had any access to technology and may be unfamiliar with common iconography.

In addition to this, no external hardware (e.g. microphones or headphones) could be expected by students. The team had to make sure all prompts and feedback were clear for students who were unfamiliar with our application without having to rely on audio.

Finally, we needed to make sure all assets, including fonts were Open License.

The Approach

We started out by coming up with the logic for each writing task based on difficulty. The writing lessons start out by instructing students how to write individual letters and numbers. Students initially see a model to copy and learn how to trace it. Once they can accomplish this, they move on to copying, and finally dictation tasks.

Tracing letters

Carrying out this section of RoboTutor included different stages. A challenge specific to this application is as designers, we cannot use any text to indicate instructions. This forced us to rely heavily on iconography, as well as different visual cues.

The screen above for word-tracing includes a model to copy, a response box, and writing area with a designated baseline. Letters are still written on an individual basis to maximize recognition using spatial distribution.


Copying digits

The screen below for number-copying includes a model to copy, a response box, and writing area with a designated baseline. Since we realized finger-input recognition could become frustrating if it constantly recognized a digit incorrectly, we decided to add a number keyboard that would be available once students have mastered all numbers.

Copying sentences

The sentence stage turned out to be the most challenging. There were different aspects to take into account:

  • Scrolling within a sentence, a gesture that might not be intuitive for first-time tablet users
  • Editing within a word, either because of incorrect recognition or incorrect input
  • Avoiding a “backspace” button that the students might not be as familiar with.
  • How to give immediate feedback
Immediate feedback

In order to provide feedback, incorrect characters are highlighted in red in the model and response. In addition to this, in cases of character omission/commission, the specific region will be highlighted in the writing area and compared to the correct character.

The screen above for sentence-tracing includes a model to copy, a response area, and writing area with a designated baseline. Several new items are introduced here: scrolling, input mode (keyboard/finger) and a “Go” button for when the student has finished.
Prompting students to take action

A hand icon pointing towards the center of the writing area shows whenever the student has remained idle too long.


My initial sketches addressed how to give feedback and how to motivate further attempts.



Model/response layout

We spent a lot of time going back and forth figuring out the best way to display the model to copy, the recognized response and the writing area.

One of the initial screen had a much less linear layout and was not as flexible.
The current version maximizes the screen space and provides the student with boundaries to differentiate between upper and lowercase letters.


Backspace vs. editing within the response

Editing a response was also something we constantly iterated on, since we could not asume all students would be familiar with a ‘backspace’ button.


The initial designs included a standard Android backspace button.


Later iterations played with editing within the response by having a ‘container’ that indicated where the next character would be displayed.


Moving forward

The project is currently under implementation and testing in Tanzania. First round of competition will take place in January 2017.