Entitled “The ethics of AI at the intersection of nutrition and behaviour change”, the new seven-step guidance aims to address safety and privacy concerns, whilst underlining AI’s potential to increase access and affordability to personalised nutrition solutions.
It also highlights the potential for AI to efficiently analyse large datasets and identify hidden patterns which may enable a shift towards preventative health solutions using nutrition.
The report concludes that the development of ethical AI solutions requires the training of systems using data representative of various cultures and preferences, utilising continued human oversight and stakeholder collaboration, whilst ensuring compliance with AI regulation.
"The reason why I am proud of this paper is that with all the frameworks that have been published on the ethics of AI, none have focused specifically on nutrition,” said Mariette Abrahams, CEO and founder of Qina. “With consumers prioritizing their health, and we all need to eat, we need to make sure that entrepreneurs creating nutrition and wellness solutions, do not underestimate the impact these digital tools have on our health, behaviours and society.
“We hope that companies will use this framework to assess internally where they are with regard to the 7 pillars and where they will need to improve or seek external help."
AI issues and opportunities
The personalised nutrition market could expand to as much as $64 billion by 2040 (UBS, 2020).
Paired with the recent surge in adoption of technologies such as wearables and the increasing implementation of AI within the health space, there is a renewed potential for the widespread application of personalised nutrition, according to Qina’s report.
However, the significant costs, limited scientific backing, substantial resources and data sets required, and lack of consumer trust represent prevalent challenges for the sector. In fact, 75% of customers reported being seriously concerned about their ability to protect the privacy of their personal health data, according to a 2022 report by the American Medical Association.
The framework
Qina’s ethical framework includes the seven interrelated principles of data—AI system, human-centric, organisation, education & training, people & planet and regulation.
Under the pillar of data, the report outlines the need for AI systems to be trained using a comprehensive and diverse dataset that is aligned with the nutritional guidelines of the target population, while continuously incorporating the most recent academic research.
“The AI solution should also train on existing behavioral data, such as insights into the eating habits, preferences and lifestyle choices of individuals to better understand and predict user behavior in real-world contexts,” it states.
The paper underlines the need for continued human agency and oversight for accurate personalised AI nutrition that involves nutritionists, data scientists and behavioural scientists.
Abrahams noted that not many companies share AI learnings and data among stakeholders, stressing their “social responsibility” to do so to improve future solutions.
The report highlights the importance of user freedom in making the final decisions that are in line with their individual values and preferences.
“A user might receive a range of meal suggestions but ultimately decides what to eat, ensuring the AI supports rather than dictates dietary decisions, an approach adopted by ZOE,” the authors explained, stressing the need to consider difference in cultures, dietary needs and preferences, and health conditions.
In addition, compliance with legal and regulatory standards such as the AI Act (introduced in December 2023) is critical for the transparency and safety of such systems, as well as compliance with The EU General Data Protection Regulation (GDPR) and Data Protection Act 2018 (DPA 2018) to ensure data protection.
For the future
The authors noted that future AI systems will require increased transparency, inclusivity and safety to ensure successful implementation as health improvement tools. They emphasised that education and training within nutrition must evolve to include design thinking, creative problem-solving and behavioural innovation.
“They should equip the next generation with the skills, tools and mindset to solve real world problems,” Abrahams said. “Practitioners get trained very scientifically which is great, but we should also learn other skills which not only make practitioners more marketable, but also help them to contribute more broadly to new product development."
“We also get trained in cognitive behavioural therapy, but this is only one aspect of behavioural science," she added. "We need practitioners to get a much better and deeper understanding of what drives behaviour and engagement especially from a digital perspective."