Artificial intelligence (AI) is quickly becoming part of how we manage our health, fitness, mental well-being, and performance. AI can create workouts, analyze information from wearable devices, suggest health strategies, support mental health services, and help athletes and performers improve. AI can save time, process large amounts of information, and help personalize programs. But there is an important question we should be asking:
Is the information AI gives us equally accurate and useful for everyone?
The answer is no. AI learns from existing information. If that information contains gaps or biases, AI will be informed by those problems too. This is especially important in health and wellness, where women, racial and ethnic minorities, and other populations have historically been underrepresented in research. Rather than assuming AI is fair and accurate, it is best to approach it as “biased until proven neutral.”
Recognizing AI bias does not mean we should stop using AI. It means we need to use it thoughtfully. AI Can Be Helpful Without Being Perfect.
Consider wearable technology. Smartwatches and fitness trackers can measure heart rate, activity, sleep, and other health information. However, some research examining wearable heart-rate technology across different skin tones has found that these devices are not as accurate on darker skin. At best, it is suggested that further research is needed to explore this issue (Koerber et al., 2022). However, the problem is that use of wearable technology has vastly expanded and is commonly used to make decisions about workouts, intensity, and sleep. The point is not that wearable technology does not work; it is that we should not assume that every technology works equally well for every person. So, this brings into questions about AI-generated exercise programs, health recommendations, mental health tools, and performance strategies.
AI + Human Intelligence
AI can analyze information quickly, but it does not know everything about the person receiving the recommendation. Imagine that AI creates an exercise program for someone based on their age, weight, and fitness goals. The program might look perfect on paper. But what if that person has limited access to equipment? What if they work two jobs? What if they have physical limitations, cultural preferences, financial concerns, or previous experiences that affect their ability to follow the program?
That is where human judgment matters. A health coach, fitness professional, mental health provider, or performance specialist can look beyond the data and consider the whole person.
AI can provide information. People provide context.
The goal should not be replacing professional expertise with AI. It should be using AI to support better human decision-making.
A Simple Three-Level Check
One way professionals can evaluate AI is by looking at it from three levels: micro, meso, and macro. In simpler terms: the person, the professional, and the system.
Micro: The Person
Start with the individual. Immediately, these questions need to be asked. Who is the individual making the decision to use AI and what is this person’s biases? Personal bias and context matter significantly. Does the recommendation make sense for the particular person that the AI is being applied to? Consider factors such as age, health, physical ability, gender, culture, available resources, goals, and life circumstances. A recommendation can be technically correct but still be wrong for the person receiving it.
Meso: The Professional
Next, consider how professionals and organizations are using AI. Is a health coach, trainer, mental health provider, or other professional reviewing the information? Or is the AI recommendation simply being accepted because a computer generated it? Professionals should be willing to question AI, adjust its recommendations, and use their own training and experience. They should have cultural competency and knowledge of the existence of AI bias and how it shows up. AI should be a tool, not the final decision-maker.
Macro: The System
Finally, look at the bigger picture. Who created the technology? What information was used to develop it? Who was included in that information? More importantly, who might have been left out? Who can be hurt or misaligned when AI is applied broadly? These questions matter because bias can enter AI systems in many ways. The National Institute of Standards and Technology (NIST) has emphasized that AI bias can come from technology, human decisions, and larger social and institutional systems (NIST, 2022).
Five Questions to Ask About AI
You don’t have to understand computer programming to think critically about AI. Before relying on an AI recommendation, ask:
- Who was represented when this technology was developed or tested?
- Could it be making assumptions that don’t fit this person?
- Has it been tested with different groups of people?
- Is a qualified person reviewing important recommendations?
- What happens if the AI gets it wrong?
These questions change the conversation from simply asking “Does this technology work?” to asking “Does it work for me or for the person I am trying to help?”
Using AI Responsibly
AI isn’t going away, and that is not a bad thing. Used responsibly, it can make health, fitness, mental health, and performance services more efficient and personalized. However, faster does not automatically mean better and personalized does not automatically mean unbiased. As AI becomes more common, people need to understand what it can do and where it can fall short. We need to question recommendations, recognize possible bias, and know when human judgment should take priority. Believing “biased until proven neutral” doesn’t mean assuming AI is bad. It means asking better questions before trusting it.
Who created it? Who was represented? Who might have been overlooked? Does this recommendation actually fit the person in front of me? These are starter questions!
The future of health and human performance should not be about choosing between artificial intelligence and human intelligence. It should be about learning how to use both responsibly.
References
Koerber, D., Khan, S., Shamsheri, T., Kirubarajan, A., & Mehta, S. (2022). Accuracy of heart rate measurement with wrist-worn wearable devices in various skin tones: A systematic review. Journal of Racial and Ethnic Health Disparities. https://doi.org/10.1007/s40615-022-01446-9
National Institute of Standards and Technology. (2022). Towards a standard for identifying and managing bias in artificial intelligence (NIST Special Publication 1270). U.S. Department of Commerce. https://doi.org/10.6028/NIST.SP.1270