Crowdsourcing Compassion: Humane AI and the Rise of Patient-Led Discovery

Listening to the Digital Whisper: Human-Centered Data Collection

PLM may have been one of the first and largest centralized OHCs, but it is certainly not the only one. OHCs like PLM are emerging as powerful user-led platforms, especially useful for patients with complex symptoms and rare diseases. Facebook groups in particular are increasingly providing a sense of community and belonging for individuals facing medical challenges; over 1.8 billion Facebook users engage in health-related groups monthly, forming over 10 million communities. ² Other platforms like Reddit and TikTok also host groups with similarly impressive numbers.

Once isolated, able to gather information only from occasional studies with limited research funding, patients with rare diseases can now easily find peers with similar conditions and compare notes. Not only do these communities provide information for individual patients, but the data contained in these platforms, when gathered in one place, helps medical professionals and researchers to see larger trends when it comes to symptoms, disease evolution, and treatment efficacy.

One recent obvious example is that of long COVID. As the pandemic swept the globe, most research focused on acute cases. Experts mapped symptoms and measured outcomes. Health care workers were focused on managing and containing the emergency, and nearly all available effort was dedicated to addressing the most severe active cases. Quickly, though, it became apparent that not everyone was recovering from COVID at the same rate, or at all. While physicians and researchers were working full-time to help people survive, there were some patients for whom symptoms lingered or new symptoms appeared. At first, most of these patients were told that their symptoms would ease and they would recover with time. But patients were beginning to find each other online to share information about ailments that would not go away. Because of these online communities, places where patients amassed a mountain of data about symptoms and helpful treatments, doctors, researchers, and even the media began to take note of the constellation of symptoms that would later be called long COVID.

Once few and far between, these patient collectives are now common digital spaces for connection, advocacy, and discovery. Their activity has already influenced clinical trials, biotech regulation, and funding policies. While health systems and academic research were not originally centered around collaboration with patients, patient and public involvement has been evolving in the United States and Europe since the early 1980s and 90s, when activists responding to the HIV pandemic began to lobby officials. ³ These advocates convinced public health authorities, including the US Food and Drug Administration in the U.S. and the newly established European Medicines Agency in Europe, that the regulatory process should address patients’ interests by fast-tracking approvals and early access to life-saving medications. This activism set a new precedent for industry collaboration with patients.

Today, stakeholders in medicine, digital health and data, medical devices, and health systems are working to better understand the value of patient involvement for driving innovative development of medications, devices, and services. ³ Patients are now frequently appointed to scientific advisory boards of major institutions and companies. In the government, programs like the European Patients’ Academy on Therapeutic Innovation, EU4Health, ⁴ and the European Patients’ Forum ⁵ require or incentivize patient involvement in product development and funding. In the United States, the FDA has adopted several measures, including its Patient-Focused Drug Development Program, to engage patients in its processes. ⁶

Patient involvement leads to a better understanding of patient needs and personalization, clear direction in early research, more efficient resource allocation, and effective trial protocols, and the industry has taken notice.

Future Directions: A Humane Model of AI-Patient Collaboration

In recent years—as it has in most industries—artificial intelligence (AI) has begun to play an ever more significant role in managing the volume of information that these OHCs produce. OHCs are decentralized and fragmented, creating pools of messy data that can be hard to collect and analyze. However, AI models can now scrape Reddit, patient forums, TikTok, and other social media platforms to detect emerging illness patterns. With these intelligent systems, patient experiences are collected from disparate corners of the web, combined with similar stories, and becoming more visible to peers, caregivers, and researchers alike.

Now that AI is no longer limited to lab-generated data, these programs can mine online platforms for patient-reported information, increasingly bringing to light concrete evidence of the insights first voiced by rare disease communities—those long overlooked by traditional medicine. AI models that scan forums and blogs are uncovering new syndromes and diagnostic clues, often ahead of the formal medical system.

Of course, innovation is never without concerns, and these forums are no exception. Some have raised ethical questions about consent, privacy, and emotional vulnerability. Behavioral health professionals have noted that while sharing with peers may be helpful for some patients, there are others who will become more anxious after reading about their peers’ experiences or may become obsessive about scouring these sites for information. There have also been issues with clinical trials in which participants have worked together on these platforms to determine who is in the treatment and placebo groups, essentially ruining objectivity and tainting the data of entire studies.

There are also tech-specific issues about AI in particular. Like other sites, PLM, for example, has incorporated an Ella AI chatbot into the homepage of its site. While Ella is clearly marked as an AI, it functions more as a recommendation engine than a conversational partner—highlighting the diversity in how AI tools engage with users. Before a new user can even register for an account, his or her activities are being monitored. In this instance, the chatbot is very clearly AI-generated, but some sites are less clear about this. In some cases, users may think they are chatting with a real human or may be unaware that their chats and posts are being mined and their data potentially shared with companies or other actors.

It is also important for users to note that many patient forums are for-profit ventures, as are most social media sites. Yes, they are providing a service for which there is clearly a great deal of demand, and often the benefits for individual users outweigh the costs. But it is important for users and researchers alike to understand that the sites they are using often have motives beyond simply sharing information. Many are supported by ad revenue and often work with their advertisers to alert users to products and services tailored to their specific conditions.

While all of these criticisms are valid to varying degrees, there are ways to temper the negative effects while still empowering patients to use technology to find community, share information, and tell their stories.

This philosophy of co-creation—where technology adapts to human needs rather than the reverse—remains vital as we move from VR to AI. More than three decades ago, when we first introduced virtual reality into behavioral health care, it quickly became clear that the technology was most effective when rooted in a human-centered approach. Patients achieved immersion more readily when the virtual environments included familiar cues. With their consent, we invited graphic artists and computer scientists into therapy sessions, enabling patients to directly shape the virtual worlds they would later use—an early, powerful example of collaborative design between clinicians, technologists, and those they aimed to serve.

In the same way, AI will be most effective (and humane) if it is designed to work with patient communities to amplify patient voices. As AI systems begin to scan online discussions, patients can become key collaborators in creating more humane, responsive health care systems and tools. By analyzing data that comes directly from patients, AI can augment human insight, centering lived experience over sterile (though controlled) laboratory data.

Sir William Osler famously emphasized the importance of listening to patients, believing that they often reveal their diagnosis through their descriptions of symptoms and experiences. He wrote, “Medicine is learned by the bedside and not in the classroom.” ⁷ In this increasingly digital world, though the bedside is now virtual, Osler’s truth still holds. Patients know their bodies best. Technology should be used to help them share and to help us listen.