The inspiration for AI-powered marking software Graide was simple. Says CEO Manjinder Kainth with a dry laugh: “I was doing a lot of teaching and marking, and just experiencing the pain that is marking. It’s quite an intense pain.”
Whilst working as a PhD student and teaching assistant at Birmingham university’s Department of Theoretical Physics, Manjinder collaborated with other students to find a solution to this ‘pain’, including fellow teacher Robert Stanyon, whose thesis was exploring the use of artificial intelligence to mark mathematical work.
The group became 6 Bit Education, who are now trialling Graide in three British universities where 96% of users say they want to carry on using the platform. Thanks to sophisticated algorithms, the system actually learns an individual assessor’s marking style, meaning they never have to mark the same answer twice. It also means the process speeds up over time – over 90% of users say the system has become significantly faster or faster than the previous month. 6 Bit Education estimate that the system can reduce grading times by an average of 89%.
This is crucial – in 2020, 31% of UK teachers worked more than 51 hours a week, 11 of which were spent on grading, and universities spend an estimated £130 per student per year on grading and feedback. Heavy marking workloads are one of the leading causes making teachers abandon the profession.
Manjinder says: “I’ve always wanted to make personal learning happen at scale. I think if we had the capacity, it would be brilliant to give every single student their own personal tutor to give them bespoke feedback, exactly understand their problems, their learning outcomes and genuinely give them a personal learning journey. And that’s why tutors are so effective, but also incredibly expensive.”
This is where Graide comes in. It aims to give teaching staff their time back, freeing them up to “be more personal” in their teaching and “keeping them at the heart of education”.
Manjinder and his team are expert coders in their own fields – he’s written one of the fastest matrix diagonalization codes to exist – but this didn’t necessarily help when trying to design a product for customers. The journey from students to business founders included learning about branding, how to build a business and customer base, and how to set up appropriate corporate structures.
Talking about the experience he says: “When you have to work together on a project, which hundreds, if not thousands, of people are going to end up actually using, understanding the nuances of user experience and trying to make the process really simple, is incredibly important.”
“We went through multiple iterations of the product with real user feedback, taking all the feedback on board, and iterating it. And then we did it again. And again, to the point where now we’ve got a pretty slick product, which is easy to use, and only takes 5–10 minutes to understand how it works.”
The hard work has paid off – in May this year 6 Bit Education received 800k from the MEIF Proof of Concept & Early Stage Fund to commercialise its product, which will allow the team to grow and add new features to the platform.
Graide is an automated process and fully designed to shift the labour from man to machine, but a big focus of its development was about preserving the organic process of completing assessments and learning from mistakes. The team knew that the main way students improve is from the feedback on their working out.
Manjinder says: “When students get questions right, that’s great, and it shows they’re improving, but students getting questions wrong is the really important part. Currently, ed tech only deals with the final answer. And that’s like seeing a symptom without understanding the history.
“If you actually care about the formative learning journey like we do, you have to be able to give feedback to the method and that’s why we developed this platform. That was at the core, and we figured if we can’t give feedback to the method, there’s no point in carrying on.”
Up to this point Graide has been aimed solely at higher education institutions, but it could potentially be even more effective in schools. Answers to university assessments are often variable and long, and and require lots of detailed feedback, but in secondary education and earlier, answers are less variable, shorter, and require less personalised feedback, which is a simpler problem for AI to solve.
An unintended upside of Graide is that there is inbuilt moderation for markers. Answers and feedback are shared across users which provides a “really nice learning opportunity”, and maintains a level of consistency in marking. It’s then much easier to spot, and correct, rogue habits in individual markers.
Of course, all of this is highly applicable to STEM subjects that have relatively objective assessments with concrete answers, but what about the arts and humanities? Are long essay teachers doomed to laborious marking for evermore?
Manjinder says no, that 6 Bit Education wants to expand into every subject, right up to the longest-form answers. He also thinks that non-STEM students are just as capable of automating processes as his team with their tech and coding backgrounds – the only difference is a lack of exposure to automation systems.
“Coding literally is just another language, but I think it has the added benefit of there being no interpretation. With english, you write a sentence, you put a comma in the wrong place, and all of a sudden the sentence says something completely different. If you put a comma in the wrong place in the code, it doesn’t work. That kind of structure is quite useful.
“It’s not this super high level stuff. I genuinely think everybody can code. For me, it’s like, english, maths, code; they’re all languages. The more the more versatile you are with those languages, the more things you can do.”
Read more: University of Birmingham alumni raise £800k for auto-grading platform