Provide a one-line summary of your solution.
Harness LLMs to recruit student interest in Math through verbal and visual problems, without relying on LLMs to do the math
What type of organization is your solution team?
For-profit, including B-Corp or similar models
What is the name of the organization that is affiliated with your solution?
Mote Technologies, Inc.
Film your elevator pitch.
What is your solution?
Please view Conker at www.conker.ai and view the Linear Relationships prototype here. See Conker's YouTube page here.
Conker is a web application for teachers to create question sets and formative assessments. It was launched in January 2023, and has so far been used by more than 220K users to create more than 800K quizzes and assessments. As well as exporting quizzes to other platforms, Conker also enables students to complete quizzes on the platform, and so far 205K quizzes have been submitted and graded to help students and educators take stock of their learning.
Conker Math is our vision for making Math practice sets and formative assessments that harness large language models (LLMs) to help to recruit interest in Math problems for Middle and High School students.
LLMs, surprisingly, can't understand and implement Math reliably. However, when provided with mathematical questions, answers & distractors, LLMs can be excellent at labor-intensive tasks including:
- Generating authentic and original mathematical word problems
- Generating character-driven stories about mathematical concepts
- Generating data & metadata for drawing graphs and other visual aids
- Generating immediately useful feedback on correct and incorrect answers
Conker Math's approach - as shown in the Linear Relations prototype - is to:
1) Build standards-aligned question formats that can be easily searched and combined into assessments by educators and/or students.
1) Create unique question sets uniquely tailored to a teacher or student's need, location and expressed preferences
2) For each question set request, to generate unique sets of correct Question / Answer / Distractors using SymPy or similar Python libraries
3) Use LLMs to develop word problems, stories and graphing data (e.g. scales, units) needed to create unique and interesting math problem sets.
4) Use modern front-end frameworks to create delightful charts and graphs, and to provide tailored assistive supports
4) Use LLMs to give feedback to students on their correct / incorrect answers and to generate new problems that help students to achieve mastery.
How will your solution impact the lives of priority Pre-K-8 learners and their educators?
"All of the area / perimeter problems are about fences and gardens. My students can't relate to that." one Math Curriculum & Instruction lead at a Texas School District told us.
Recruiting the interest of Math learners is essential if we want all learners to overcome their personal and community challenges and to achieve their potential. Post pandemic, the performance gap for Black and Latino learners is especially pronounced in Math.
Generative AI offers the possibility to optimize the relevance of math assessment materials to help students engage with the concepts, through characters, stories and examples that they can relate to.
How are you and your team (if you have one) well-positioned to deliver this solution?
We are a small team, consisting of two former Texas based educators and three product builders. We have been on a journey exploring the power and the limits of technology to make a positive impact for educators and learners. We launched our first Edtech product (Mote) in March 2020, then saw it grow to hundreds of thousands of users in a matter of months. We began work on 'Conker' in October 2022, before ChatGPT's launch, because we saw the potentially transformative impact of generative AI within education.
Since then, we have seen both the potential of generative AI and its limitations. We have seen the demand for better questions and assessments from Math educators, and realized the inadequacy of generative AI to meet that need. There is no 'out of the box' solution for Math with Generative AI - much more work is needed to build Math-specific assessments that are powered by Generative AI.
We spent three months working on the prototype of Linear Relations, with a small dedicated team, and learned a lot about the challenges of harnessing LLMs reliably and scalably.
We then worked on a follow-up project with a leading Edtech publisher to build a 'Proof of Concept' for implementing a similar Math experience within their application, which serves millions of users around the world.
We have been gathering feedback from Math teachers and Math Curriculum & Instructional specialists to validate that there is a real need here.
As a portfolio member of the New Schools Venture Fund, we have been part of a diverse community of practitioners working on transformative educational solutions. Through New Schools, we have better understood the challenges of underserved communities however we are still early in this journey.
Which dimension(s) of the challenge does your solution most closely address?
Which types of learners (and their educatiors) is your solution targeted to address?
What is your solution’s stage of development?
PilotIn what city, town, or region is your solution team headquartered?
London, UKIs your solution currently active (i.e. being piloted, reaching learners or educators) within the US?
Yes
In which US states do you currently operate?
We have Conker users in all 50 states and District of Columbia. We have piloted with two school districts in Texas, and have a paying school customer in Florida. Texas is our strongest state, partly because we have focused on the TEKS standards and STAAR testing, and partly because we have strong relationships with several districts there.
Who is the Team Lead for your solution?
Will Jackson
What makes your solution innovative?
Conker is far from unique in harnessing LLMs to generate assessments for educators, however we hear from many of our partners that they have tried all the other generic tools, and prefer Conker because of the quality and the attention to detail in our product. In particular:
- Modeling question formats based on Texas' STAAR testing, which is the first example of fully online state testing.
- Offering 'Read Aloud' accommodations for students
- Supporting logged out and logged in responses
- Grading and adding feedback for every correct and incorrect answer
Conker's approach to Math is very unique and innovative in the market, combining Python-based modeling of math questions and answers with LLM-based story-telling and word problem generation.
While the prototype is simple today, the potential is clear: students can create their own unique characters who can embark on journeys through mathematical content, while still covering the same curriculum standards as their class mates.
Describe the core AI and other technology that powers your solution.
Conker has been built using Generative AI, in particular using OpenAI's GPT 3.5 and GPT 4 models.
Our full tech stack comprises a Python-based core services infrastructure, combined with a Node.js front end, hosted with Vercel. We have implemented various front end frameworks for graphing.
How do you know that this technology works?
The product can be tested freely as follows:
Conker: https://app.conker.ai
Conker Math (Linear Relationships prototype): here
What is your approach to ensuring equity and combating bias in your implementation of AI?
We take equity and bias seriously, and know that they are of central importance to our users and customers.
With respect to equity, we have taken steps to ensure that we reflect the diversity of our student user base through
- Integrated 'read aloud' functionality on every Conker quiz
- Support for developing assessments in a wide variety of languages
- In our 'Quiz bank', offering the Spanish language versions of the TEKS standards where available
- Soliciting product design input from a diverse cross section of schools and educators
We have not had any issues of bias reported to us to date, to the best of my knowledge, and we receive plenty of other feedback on Conker. We are very thoughtful when developing prompts to provide appropriate context for the model.
We look to our AI technology provider to implement safeguards against bias, and would welcome other suggestions for how to test for bias more robustly.
How many people work on your solution team?
Our full time team covering Mote and Conker is 5, namely:
CEO
CTO (back end engineering)
Front End Engineer
Partnerships Manager (sales)
Customer Success Manager
How long have you been working on your solution?
1.5 years in total.
If your solution has a website or an app, provide the links here:
www.conker.ai
What is your plan for being pilot ready (if not already) within the next year, and what evidence can you provide that you are on track to meet your goals?
We are already in market with Conker. If we win this prize, we would invest in doubling down on Conker Math, with regular (monthly) releases as we extended support into new curriculum / standards areas.
What are your plans to ensure your solution is available, accessible, and affordable to priority learners at scale?
Conker's pricing of around $60 / year for an educator license is discounted at scale, and we would target around $2 / student / year at scale.
Why are you applying to the Learner//Meets//Future Challenge?
Developing Conker into a fully featured product that addresses students' need for more relevant, authentic and personal math assessments will be a multi-year journey that will likely need significant investment ($1-2m) to realize.
In current market conditions, it's hard to see that investment coming from private investors. We have seen that schools are slower than ever to invest in innovative solutions, given the budgetary constraints that they are facing. The Bill & Melinda Gates foundation would be our ideal long term partner in scaling Conker's ambitious plans, and the foundations support as a cornerstone could perhaps help us to bring in wider support from other financial and social investors.
In which of the following areas do you most need partners or support?
Solution Team
-
Will Jackson Mote Technologies, Inc.
to Top
What is the name of your solution?
Conker