The Best of Both Worlds: Hybrid human expert + AI coaching

Collaborative cases: Where human machine synergy outcomes outperform either

There was a time when a monk would advise you about inner peace, and how you can achieve it once you’ve got the bare necessities of food, clothing and shelter sorted. Not today, no amount of meditation can pacify you once you’re offline or your phone runs out of charge. In the connected era, there will always be those rare few who prefer to disconnect, but it’s progressively difficult for the average internet user.  Have you noticed the transition in your own behaviour over the past 20 years? How we went from being able to have instant recall from memory about the capital of a country or a restaurant name or all possible drug interactions - to having the incessant need to look up everything with an internet search. We lean on our devices as crutches to go by our day. Seems pessimistic at first glance, but don’t get me wrong, this is a good thing. Neural scaling laws indicate that the more data we train with, the better the AI beast roars. After all, today, the narrative is data is the new oil – which is why all foundation model trainers are exploring new data sources. Interesting how we’ve crawled the whole internet and have now run out of accessible data to train models! Note the stress on accessible, not all human knowledge is easily tokenizable text. (Books, Movies, Videos, Images, Enterprise data, and most importantly - Wisdom)

Which takes me to my case – we will reach the ceiling of individual model performance very soon, till we develop better data preparation techniques. However, the next leg of intelligence in the short term will not flow from models that can utter better, but more from models that can complement humans better -which is why people chose Perplexity over Google, Claude over GPT for code completion and so on. Our new crutch after the internet and the smartphone will be decision enhancing workflows using AI. And that is where the constraints of our human intuition lift and get adapted to a new paradigm. The synergistic adaptive intuition paradigm. In a past life, when I worked on computer vision products, there was an abundance of  studies that proved that the combination of a radiologist and a computer aided detection system was far superior in sensitivity and specificity than the radiologist alone. The US FDA has seen a progressive increase from 2016 in clearances with over 660 approvals of algorithms in comparison to only 30 in the previous 2 decades. No prizes for guessing most of the 660 are deep learning algorithms.

Not just detection, but many AI companies in imaging have focussed on the efficiency play – either by prioritising the abnormal scans the radiologist should read, or in case of emergencies, prioritising the most critical ones. In addition to time savings and accuracy improvements, for an algorithm to be successful, it also needs to demonstrate a benefit in terms of cost and health outcomes than standard of care. Then it becomes the new standard. The collaborative case exists on the intervention side too. Robotic surgical systems have led to progressive decrease in post-op complications, newer generation LASIK systems have led to faster operative turnaround and return to day-to-day life. Perhaps adoption has been slower with AI owing to the black box problem. But we shall overcome. And with the width and depth of machine intelligence available today, I still feel there’s scope to improve the outcomes through human and machine synergy.

The Human Element: Why Empathy and Understanding Can't Be Automated

 Empathy, emotional intelligence, and the ability to understand and respond to the unique needs and concerns of each patient are qualities that cannot be fully automated. It’s not a statement about the possibility and feasibility, I comment hoping that most of us see the wisdom in not going down this path.

The importance of the human element becomes lucid when you consider a patient who has just received a devastating diagnosis. An AI chatbot today can provide accurate information about the condition, treatment options, and prognosis. However, it lacks the ability to empathise with the patient's fear, grief, and uncertainty. A doctor today offers not just information, but also compassion, reassurance, and a listening ear. They can pick up on subtle cues in the patient's tone, body language, and unspoken concerns that a machine may miss. 

Moreover, healthcare often involves complex decision-making that goes beyond medical facts. Patients bring their own values, preferences, and life circumstances to the table. A skilled and empathetic human provider engages in shared decision-making, helping the patient weigh the risks and benefits of different options in the context of their unique situation. They can provide guidance and support that is tailored to the individual, taking into account not just their medical needs but also their emotional, social, and spiritual well-being. The role of the AI here is better suited in expanding and adapting the provider’s intuition to deal with the case.

We tend to think in silos and in our own individual capacities. We emote in our own silos and individual nuances. We generate on average about 1.7 MB of data per second on the internet. However, like I mentioned earlier not all of it can be crawled. And exactly like that, our emotions, empathy and humanity can’t be crawled. All the emotional intelligence work being explored today, I fear grossly averages population level data, and misses individual nuances. And when it comes to talking, teaching, coaching, explaining a diagnosis, guiding on a treatment plan, as developers it would be prudent for us to not code for empathy. We must recognize the inherent limitations of AI in replicating emotional intelligence. While AI can augment and support healthcare in many ways, it cannot and should not attempt to replace the critical human element of care.

Care at Scale: Leveraging AI for Population health management 

In a previous post, we explored the idea that with AI, “We wouldn't be operating in healthcare as interventional and therapeutic specialists. We would be operating in healthcare as preventive specialists." We can build fabulous models to identify trends, predict risks, and optimize interventions to improve the health of entire communities. If deployed, imagine a future where algorithms continuously monitor and analyze data from electronic health records, claims databases, social determinants of health, and even wearable devices and social media. By identifying patterns and risk factors, these systems could alert healthcare providers and public health officials to emerging threats, such as disease outbreaks or clusters of chronic conditions in certain populations.

AI could be a powerful tool in realizing this vision of proactive, preventive care at scale. For example, an AI system might identify a group of patients at high risk for developing diabetes based on factors like age, family history, and lifestyle habits. Healthcare organizations could then proactively reach out to these individuals with personalized prevention programs. At an even broader level, AI could help inform public health policies and interventions. By analyzing data on social determinants of health, such as housing, education, and access to healthy food, AI could highlight areas of need and guide community-level investments to address health disparities. One might be quick to retort with the ethical considerations around privacy, bias, and equity that can deter this approach. But the empathy of human decision makers in the loop is what can differentiate these hybrid systems and make them transparent, accountable, and aligned with our values as a society.

This synergistic adaptive intuition paradigm I proposed earlier – where human expertise is augmented and enhanced by AI insights – is particularly relevant in the context of population health. Epidemiologists, public health experts, and community health workers could leverage AI tools to inform and guide their efforts, while still bringing their deep understanding of local contexts and relationships to bear. By combining the strengths of human wisdom and machine intelligence, we have the potential to create a healthcare system that is both deeply personal and broadly effective – one that delivers the right interventions to the right populations at the right time.

 At Fitterfly, we believe that by leveraging technology in service of human connection and care, we can make a meaningful difference in the health of individuals and populations alike for Chronic diseases. By empowering our coaches to manage our subscribing members better with personalised insights, proactive recommendations, CGM analytics, PGR scores, and other upcoming tools, we’re doing our part to strengthening our Adaptive Intuition.