Indoni Developers is a non-profit organisation for women in ICT. It started last year as a WhatsApp group to teach women to code and recently held its inaugural developers conference, bringing together attendees and speakers from across the continent for a day of community building and skills development.
Sewagodimo Matlapeng is a software engineer at financial software firm Yoco and a co-founder of Indoni Developers. She said the positive response to the coding classes showed that there is a need to create a space where female developers can learn and share their successes.
Pelonomi Moiloa, a data scientist at Nedbank, delivered a presentation on how to avoid embedding biases into algorithms. Moiloa is also a co-founder of Code Kamoso, a coding academy.
“My job is to put machines out there into the world, so I’m interested in the unintended
consequences that these machines can have.” Moiloa says the time she spent studying in
Japan, relying heavily on translator applications, showed her how much power technology has.
“These machines also empower all sorts of other tech: intelligent Web searching, AI-powered
chatbots, autocomplete; it’s used in education, in courts…but word-embedding technologies
learn from history, capturing hundreds of years of discrimination.”
Protecting machines from us
She gives some examples: “Google shows men higher paying job ads and Amazon had to scrap its sexist AI recruiter.”
She says the word2vec algorithm, which is used in natural language processing (NLP), will change a sentence from ‘she is a doctor’, ‘he is a nurse’, to ‘he is a doctor’, ‘she is a nurse’, because it’s learned to associate men with ‘doctor’ and women with ‘nurse’.
She also mentions the COMPAS (Correctional Offender Management Profiling for Alternative Sanctions) algorithm in the United States, which is used to predict the likelihood of recidivism.
“When defendants are booked in jail, they respond to a COMPAS questionnaire. Their answers are then fed into the software to generate several scores including predicting the risk of reoffending. Black defendants are more often rated ‘high-risk’ than white defendants and are thus more likely to receive harsher punishment.”
Moiloa was asked how alternative datasets can be sourced to avoid such biases. “It’s quite tricky because datasets are primarily created in places where they have the resources to create them; this needs a lot of money and skilled individuals. But South Africa has an NLP project for processing African languages, the Masakhane Project.
Edtech
A discussion was also held on the changing landscape of education in 2020. It was facilitated by Nonnie Ngcobo, a software developer at Reflective Learning.
The conversation examined the ways to improve access to information in order to improve schoolchildren’s lives. “Access to technology is no longer just a ‘nice to have’ because we’re moving into an era where automation will take up many jobs and leave many unemployed. Considering the digital divide we’re still dealing with, who will participate in this era? And what will everyone else do?” asks Ngcobo.
One of the solutions Ngcobo suggests to bridge the divide is through personalised education. ”The benefits of personalised learning go beyond improving education for learners; it creates a better experience for teachers too – well written algorithms can give teachers more in-depth info on their students’ progress. But it also isn’t just about personalising each student’s learning, it’s about giving them access to information, helping them experience the world outside their surroundings.”
She mentions edtech company Siyavula that offers free and open-source online maths and science textbooks. Its paid service, Siyavula Practice, allows learners to answer curriculum-aligned questions. When a learner has mastered a certain level (as opposed to a grade), they move up to the next level, at their own pace.
Other suggestions from those attending the conference included getting basics such as improving connectivity, promoting self-study through online sources such as Udemy and YouTube, and persuading corporate SA to play a bigger role by donating resources such as unused computers.
Matlapeng thinks teachers should also be empowered through edtech and made more comfortable using technology so that it isn’t up to the learners to ‘lift themselves up’.
“If teachers know how to make the most of technology as a tool, their students can also be empowered," she says.
Share