Guides

TWC Health Notes Issue 2

Sometimes the most interesting part of a story isn’t the original discovery but the question it leaves unanswered. This issue brings together ten subjects that caught our attention, from a horse that fooled scientists to a car that reported on its driver. Some are serious, some are surprising, and one involves the number 17.

1. AI Can Answer Medical Questions. But Can Patients Use Those Answers?

A study published in Nature Medicine examined what happened when 1,298 people used AI chatbots to investigate possible medical conditions. When tested independently, the AI systems identified conditions correctly in nearly 95% of cases. But people using those same systems correctly identified relevant conditions less than 35% of the time, performing no better than the comparison group. That’s a remarkable difference. It suggests that an AI system’s medical knowledge doesn’t automatically translate into better decisions for the person using it. The quality of the questions, the information provided, and the person’s ability to interpret the response all become important. As AI becomes increasingly available for health questions, perhaps we should be measuring not only how well the technology answers, but how well people can use those answers.

2. Your Car May Know More About You Than You Realize

Most drivers understand that modern vehicles collect information, but how many expect their driving habits to influence their insurance premiums without their knowledge? On October 29, 2026, a coalition of state attorneys general announced a $220 million settlement with General Motors over the sale of driving and location data to brokers, who subsequently provided information to insurers. The data included driving times, hard braking, acceleration and speeding. The controversy raises a broader question: when we purchase a connected product, how much control do we actually retain over the information it collects? Cars are only one example. Smart televisions, wearable devices, mobile applications and household technology all create opportunities for information to travel well beyond its original purpose.

3. The Horse That Outsmarted the Experts, Without Doing Any Math

In the early 1900s, a horse named Clever Hans astonished audiences by apparently solving arithmetic problems. He answered questions by tapping his hoof, and even experts struggled to explain his abilities. Psychologist Oskar Pfungst eventually designed experiments that revealed something unexpected: Hans was responding to subtle, unintentional cues from people who knew the answers. When those cues were unavailable, his apparent mathematical abilities disappeared. What makes the story remarkable isn’t simply that observers were fooled. It’s that Pfungst recognized the need to test a completely different explanation. More than a century later, Clever Hans remains an important reminder that a convincing performance doesn’t necessarily demonstrate the ability we think we’re observing.

4. Why Does AI Seem to Like the Number 17?

Ask an AI to choose a random number between 1 and 30, and you might receive 17. It feels reasonably random, doesn’t it? It’s not an obvious endpoint, a round number, or a simple multiple of five. But that raises an entertaining question: is the AI selecting a number randomly, or selecting one that resembles what humans consider random? Language models generate responses using learned patterns, and those patterns can produce preferences that aren’t the same as a uniform random-number generator. Of course, one answer proves nothing. Try asking several models to choose numbers repeatedly, record their responses, and see whether certain numbers appear disproportionately often. It won’t be a scientific experiment without appropriate controls, but it could make for an interesting afternoon.

5. When Healthcare Eligibility Becomes a Paperwork Problem

Healthcare coverage is often discussed in terms of who qualifies and who doesn’t. But there’s another question: what happens when a person qualifies for assistance yet cannot navigate the administrative requirements needed to demonstrate eligibility? We recently examined concerns about Medicaid exemptions involving frailty, disability and other circumstances that may require documentation. For someone managing chronic illness, limited mobility or cognitive difficulties, gathering records and responding to notices can become an additional burden. Rules designed to distinguish eligibility categories may be reasonable in principle, yet the procedures used to administer them can create obstacles for the people those rules are intended to protect. Access to healthcare depends not only on eligibility, but also on whether people can successfully navigate the system.

6. Should Police Be the First Responders to Mental Health Emergencies?

When someone experiences a psychiatric crisis, a call for help may bring police officers rather than mental health professionals. Some communities have developed mobile crisis teams that respond with specially trained healthcare or behavioral health personnel, sometimes without police involvement when the situation can be handled safely. The idea seems straightforward, but making it work requires dependable funding, trained staff, coordination with emergency services and places where people can receive continuing care. The issue is particularly interesting in Hawaiʻi, where geographic separation and workforce limitations can complicate service delivery. The question isn’t simply whether police or clinicians should respond. It’s how communities can design a system that sends the appropriate help, at the appropriate time, and provides somewhere for that help to lead.

7. A Seizure Doesn’t Always Mean Epilepsy

The terms seizure and epilepsy are sometimes used interchangeably, but they describe different things. A seizure is an episode of abnormal electrical activity in the brain that may produce changes in movement, awareness, sensation or behavior. Epilepsy describes a neurological condition involving an enduring predisposition to epileptic seizures. Someone may experience a seizure because of an acute medical problem, such as severely low blood glucose or a major electrolyte disturbance, without necessarily having epilepsy. Conversely, epilepsy can be diagnosed under specific circumstances even before a person has experienced two unprovoked seizures. Understanding the difference matters because the evaluation, treatment and implications can be quite different. It’s a useful example of how a familiar medical term can conceal an important distinction.

8. Is AI Helping Us Think, or Helping Us Avoid Thinking?

One of the most interesting questions surrounding artificial intelligence concerns what happens to our own reasoning when a machine can produce answers almost instantly. AI can help us compare possibilities, identify missing information and explore unfamiliar subjects. But it can also make it tempting to accept a polished explanation without examining how the conclusion was reached. The distinction may depend less on the technology than on how we use it. Asking AI to challenge an assumption, explain competing interpretations or identify weaknesses in an argument can stimulate thinking. Asking it to make every decision may do the opposite. Perhaps the important skill in the AI era won’t be writing increasingly clever prompts, but recognizing when we should continue thinking for ourselves.

9. Can AI Make Hiring Decisions Fairer, or Simply Hide the Bias?

Companies increasingly use automated systems to screen applications, evaluate qualifications and identify promising candidates. Such tools may process information faster and more consistently than human recruiters, but consistency doesn’t guarantee fairness. If a system learns from historical employment decisions, it may reproduce patterns of exclusion that existed in those decisions. Even apparently neutral criteria can disadvantage particular groups when they correlate with characteristics unrelated to job performance. There’s also the question of accountability: when a qualified applicant is rejected by an automated system, can anyone explain why? AI hiring isn’t inherently unfair, but the promise of objectivity deserves scrutiny. A process can be technologically sophisticated and still require careful human judgment.

10. When AI Reviews Scientific Research, Who Reviews the AI?

Peer review is intended to help identify weaknesses in scientific work before publication. Increasingly, AI tools are being used to assist with tasks such as checking language, summarizing arguments and suggesting potential criticisms. Those capabilities could be useful, especially when reviewers face demanding workloads. But there is an important distinction between identifying a possible concern and establishing that the concern is valid. An AI-generated criticism may sound authoritative while misunderstanding the research design or overlooking information already provided by the authors. If reviewers accept such comments without checking them, the technology could add noise rather than improve scientific quality. The larger issue is accountability: scientific criticism should be judged by its evidence and reasoning, regardless of whether it originated with a person or a machine.

A Final Thought

These stories have something in common: they challenge the assumption that what appears obvious is necessarily correct. A horse that seems to calculate, a medical chatbot that seems knowledgeable, a hiring system that seems objective, or a car that seems to serve only its owner can each tell a more complicated story.

Curiosity begins when we recognize that there may be another explanation worth exploring.

Until next time,

Marsha
The Wellness Center for Healthy Living