Mariette Abrahams, spoke about the ‘next wave of challenges’ for personalised nutrition at the Education Programme at Vitafoods Geneve last week and warned that personal data collection strategies must be improved if personalised nutrition is to be seen as anything other than a fad.
“We’ve heard that data is the new currency but I say that data is the new controversy," she warned.
“Did you know the accuracy of these health tracking devices is, on average, just 20-50%? So the quality of this data is low."
She told the story of Joy Buolamwini, the Ghanaian-American computer scientist who famously found that a computer's facial recognition programme did not recognise her face because of her dark skin.
Abrahamas explains how this story hints at some of the dangers of allowing algorithms to determine our health needs.
“If we are training machine algorithms on data that’s already biased we will come out with biased outcomes.
“We have to use high quality data to not worsen inequalities in society or innovation could increase the inequality we see – health inequality resulting from negative socioeconomic circumstances is a substantial economic issue.”
The three Q’s of data
Abrahamas explained that there are three important Q’s to question in data collection: quantity, quality and quest.
To test how well current personalised nutrition companies has questioned these Q’s she asked 12 companies about their data collection protocols.
She found that only four of them had policies in place to identify gaps in their data or to identify bias and said ‘many’ of them did not think about how socioeconomic status or ethnic group was affecting data.
Data analytics, data science, bias identification and diversity and inclusion training were only offered, respectively, at one of the companies questioned.
“We need to improve training within companies to make sure those building products that impact health are well educated and trained in these areas," Abrahams stated.
“To companies starting out in this field I would say they must develop a data strategy which includes a data ethics audit and up-skill your team so they know and understand the importance of nutrition.
“Engage with users and stakeholders at an early stage and validate products using new research tools.
“This way you can show that personalised nutrition is not just a fad.”