Gadget

Do the maths: algorithms and tech work

South African edtech start-up Siyavula uses algorithms and interleaved learning to get maths and science learning and practice to stick, writes RUSSEL SOUTHWARD.

Most African edtech start-ups use technology as a delivery and distribution channel. South African start-up Siyavula is different because it uses a combination of technology and best education practice to improve what students remember when doing maths and science.

CEO Mark Horner and Megan Beckett, Learning Research, Design and Analytics, at the company, provided some great insights during a discussion.

At Siyavula’s core there are textbooks for maths and physical and life sciences for grades 10-12 in South Africa. It operates a hybrid of text and print and the metrics are impressive. On the physical side, it prints 10 million textbooks that go to 25,000 schools and are used by 4.5 million learners. On the digital side, its site has 6.3 million users who have had 11.2 million sessions. This is clearly no pilot project.

Although Horner is perfectly happy with this hybrid delivery approach, he still thinks there is a tantalizing prospect for change: ”If all books were not distributed in print format, we could finance a device and give each pupil one,” he says.

“We have a catalogue of openly licensed textbooks and make both print and e-textbooks available. These are all available as PDFs on our hub. Two versions are available: one is adaptable and the other under a Creative Commons ND licence. Everything available in three versions: print, PDF and HTML. In English and Afrikaans with teachers guides in all formats.

“The core content is the same but only the production and distribution costs vary. Access to textbooks is a basic requirement, but the format in which you read that book makes no difference to your learning.

“We wanted to address learning efficiency for maths and science. What makes the difference in Maths and Science is mastery-based skills. We need interleaved practice to rewire the neural pathways. If you do this, your score goes up versus blocked practice. So we built a tool to do this”.

For those not in the education field, the significance of all this may not be immediately apparent.

Mastery-based learning consists of three elements: acquiring the component skills, practicing the skills acquired and integrating them and knowing when to apply them. The latter is particularly important because where the learner has just memorized things in a narrow example, they may not have completely understood the concept or concepts that might be used in several different ways.

The starting point, as Megan Beckett says, is to ask what is the most effective form of learning: ”How do we best learn? Traditionally in the classroom, you listen to the teacher or at University you’re given a lecture or asked to read something. You don’t absorb much information through that. Active learning is where you practice by doing something and this and peer-to-peer learning are both very effective”.

According to Beckett, the fault of the education system is that it’s very content-focused and teachers are often rushing through the content in the curriculum to complete things: ”You need revisit things to solidify them in (the students’) long-term memories”.

Siyavula’s online offer allows both teachers and students to revisit what they’ve learned in a structured way and address the gaps in their understanding:”We are using algorithms to determine the level of mastery through what the student got right or wrong”. Through this process they create the “optimal cognitive load”.

So what does that mean?

”You want the work to be not too difficult but not too easy,” says Beckett. “You need to maintain learners getting things right 70% of the time. If the mark was less, weak learners might be completely overwhelmed. You need to be able to give time to develop component skills. This can also be applied to the teaching domain. They are challenged to the level that meets their prior knowledge. When the learner gets things wrong, how frequently have they seen it? How long ago? Do we need to show it to them again?”

This is where the ideas of blocked and interleaved learning comes in. In traditional education practice, the teacher designs work to “drill and kill”: the students get to memorise and sometimes understand a particular aspect of a subject. You have the students practice until they’ve mastered that particular piece of learning.

This is called blocked practice and, by contrast, interleaved learning practice goes back over several different aspects of what has been taught previously in a connected way to bring out underlying concepts.

”Doing blocked pactice in this way you do well at the time but when you have to recall for a test much later, you do much less well. We teach several concepts at once and test different things. The students don’t perform well in the practice sessions but they retain what they have learnt for longer. After each time it doesn’t rely on same thing, it’s strengthening your neural pathways because you have to recall new things.”

The tech tricks in the online offer are two-fold. Firstly, there is an adaptive engine that gauges the learner’s level of skills and sequences their learning and secondly there are generative question items that can be used in many different forms over and over again, says Beckett.

“It’s the way we design our questions. We don’t design each question as a one-off. Each question is a template from which you can generate many versions of the same question. The learner has access to an unlimited number of questions. The model solution adapts to the questions shown. You can do the same question again and you will get a different context and different numbers.

“The generative question items can build out a massive bank of questions to revise or practice on. It’s not just multiple choice and learners get feedback immediately with advice on how to improve”.

The online offer can be used in a variety of settings – the classroom, with a tutor and for home schooling – and you can say go practice. The students are expected to work on paper but type in their answers to the device they’re using.

Horner says: ”It’s about creating a growth mindset, filling in bars to show progress. We try to get agency from the learners. Teachers can use it in the classroom but you can also use it as an individual.” 25% of students are sufficiently self-motivated to start practicing on their own.

Some teachers in rural areas in South Africa only teach 100 days a year, a problem that is widely found in other parts of Africa, says Horner: ”Schools are not the ideal drivers in a rural context. Getting to learners and parents directly is key.”

Delivery is designed to work across a range of devices including a basic feature phone, a tablet, a laptop or a computer. Vodacom and MTN in South Africa have zero rated the service to encourage take-up: “You create greater impact if you cater for existing devices. The app we want to target is the web browser.”

So what’s the business model? There are two types of customers, institutional (the schools) and individual (learners and parents). For learners and parents, it costs R599 ($44.20) for one subject and R999 (US$73.72) for both subjects. For schools, it varies between R120-R300 (US$8.85-US$22.13) per learner per subject per year for 200 individual subscriptions or more. You can pay using m-money or a credit card. For individual subscriptions, it is looking for retail partnerships with banks and mobile operators.

This compares to professional tutoring companies charging over R1000 per month per subject; ad­hoc tutoring which costs around R200 per hour; and international online services and locally ­licensed versions which are are all over R1000 per year.

“Lower income households have no effective options,” says Horner.

25 000 students at more than 60 schools have signed up for this practice service and there are 75 000 sponsored learners at 285 schools. Sponsorship comes from organisations like the Vodacom Foundation.

“On the sponsored schools, you can sign up for a trial and if you meet the learning targets, you can get sponsorship,” says Horner. “We’re looking towards (getting funding) from something like a social impact fund.”

What else do they know about their users? 70% live with their biological mothers; 95% can access a cellphone at home and 68% have family monthly expenditure of less than R5 000.

What about expansion into the rest of Africa? It emerges that 40 000 to 50 000 Kenyans are already using the site.

”We would like to do this through the banks and mobile community. There is a huge interest from some banks for something that is more than just a banking product.”

Siyavula’s shareholders are Omidyar, the Shuttleworth Foundation and PSG Group, a local equity fund.

* Russell Southwood is editor of Smart Monkey TV.  To subscribe to its web TV channel, visit http://www.youtube.com/user/SmartMonkeyTV/videos

Exit mobile version