Khensani is the first African woman to earn a PhD in applied data science, focusing on AI algorithmic fairness.
“A special shoutout to my UJ family! Thank you for your unwavering support! The education system needs more people like you!” Khensani wrote on her LinkedIn page after achieving this momentous and history-breaking milestone of becoming the first African woman to obtain a PhD in applied data science: AI algorithmic fairness from the University of Johannesburg.
To get a better understanding of what led to this achievement, it’s appropriate to learn where it all began. Khensani had no idea what academic field she’d pursue after matric, but some guidance from her siblings pointed her in the right direction.
“Getting into the technology sector largely had to do with my brother, Honey,” Khensani explains. “He recommended that I pursue a Bcom degree in informatics and information systems after completing my matric, which is what I then went for, and later fell in love with! After that, I did my masters in IT, followed by my current PhD in applied data science,” she adds.
Going rural
She’s a rural girl through and through, and calls Limpopo home. Khensani points out that coming from her particular neck of the woods made attaining her current qualifications even harder. “Coming from a village like mine, there were no women in the field that I could look up to as role models and aspire to become one day,” she notes. “I was just fortunate to have had a brother who was able to steer me in that direction.”
Khensani describes her upbringing as a very humble one in a remote rural village in Limpopo. “It was a place where I wasn’t exposed to much; when I say rural, I mean there were no roads, water, electricity or TVs! – at least, around the time I was growing up.”
Although Khensani didn’t have any role models from a professional perspective, her parents, she highlights, really shaped the woman she is today. Both were educators in her village, and made education a top priority in the Xivuri household.
“It was initially my father’s dream for me to pursue a PhD. However, the more I delved deeper into the topic and the journey, it quickly became my dream too. My parents have had a profound impact on shaping who I am today, and being the youngest of seven children, I was quite spoiled,” she says.
Khensani received ample attention as a child, coupled with a strong work ethic and a drive to succeed being instilled in her. “[I was raised] to strive to be something and to ‘break out’ of the village, an environment that equally truly shaped my character,” she adds.
Juggling work and school
However, the road to academic excellence was an easy one for Khensani to travel. She reflects on her biggest obstacles during her master’s and PhD studies.
“The main challenge I encountered during my studies was balancing school with my full-time job,” she says. “The beginning of the PhD journey was particularly tough, as my submissions always came back with red marks from my supervisor. However, I persevered, and it gradually improved as I started seeing growth, and found the motivation to keep pushing forward.”
That little voice in her head telling her she could do it was supported by an actual ecosystem of people wanting the best for her, with Khensani’s parents and siblings being at the centre of that support structure.
At work, Khensani’s manager played an amazing role as a supportive leader. Her supervisor believed in her and consistently pushed Khensani to her limits. “The team at the Centre for Applied Data Science at UJ has also been incredibly supportive,” she says.
AI algorithmic fairness
Khensani’s PhD thesis focuses on the development of a process model for algorithmic fairness using the Habermasian approach. She highlights the significant increase in the use of AI in organisations, accompanied by a growing awareness of bias and its profound impact on society, which inspired her thesis.
“My thesis developed a process model for ensuring algorithmic fairness in AI and stressed the importance of close collaboration with all stakeholders, including society, throughout the AI development process. The study also provides practical guidelines to identify and mitigate bias before deploying AI models. These guidelines include:
- Governance measures – the establishment of AI regulations, clear and progressive AI policies, and governance structures, and holding organisations accountable for bias in their AI models.
- Social measures – fostering collaborations between AI managers, developers, and impacted societies.
- Technical measures – conducting data reviews and validations to ensure that the data used for model training is representative, auditing models, and testing them over time for bias.
- Training and development measures – training developers on AI ethics and ensuring explainability and transparency in AI systems.
“Pretty impressive stuff for a couch-potato,” she jokes. Throw in a glass of wine, family and some laughs, and that’s all Khensani needs to unwind and take some pressure off.
