Rethinking Informed Consent for AI Trials, an Autonomous Oxygen Device, and Canada’s Physician-Led Push for Safer Social Media

Digital health publishing rarely delivers three stories in a single week that touch the same nerve from such different angles, but JMIR Publications has done exactly that with its latest batch of News and Perspectives features. One examines how informed consent must be rebuilt from the ground up for clinical trials that involve artificial intelligence. Another follows the Canadian Medical Association as it campaigns for federal legislation aimed at protecting children on social media. The third explores a closed-loop device that could take one of the most repetitive tasks in hospital care off the hands of already stretched clinicians: adjusting supplemental oxygen. Taken together, the three pieces sketch a field wrestling with a common question, namely how to keep human judgment, autonomy, and accountability at the center of systems that increasingly run themselves.

The first feature, written by JMIR Correspondent and researcher Hankun Su, takes aim at a document that most patients sign without a second thought and that most researchers treat as a legal formality. In “Informed Consent in the AI Age: Why and How Documentation Must Change,” Su argues that the traditional consent framework rests on two assumptions that no longer hold when artificial intelligence enters a clinical trial. The first assumption is that the risks to patient safety are static, fixed at the moment the protocol is written and fully knowable in advance. The second is that if any gaps in a participant’s understanding remain, a clinician will be able to fill them in during a conversation. AI undermines both assumptions at once, because algorithmic systems can behave in ways their own developers cannot fully predict and because the clinicians supervising an AI-driven trial may themselves be unable to explain why the system produced a particular output.

Su’s response is not to abandon informed consent but to redesign it for the world as it actually functions. He and his colleagues developed a framework for AI-related clinical trials built on three pillars. The first mandates transparency and disclosure, requiring that the role of the AI system, its limitations, and its failure modes be spelled out in language participants can actually use. The second applies tiered standards calibrated to risk, so that a trial involving a low-stakes algorithm does not demand the same disclosure apparatus as one in which the algorithm directly influences treatment decisions. The third employs dynamic consent mechanisms, allowing the document to evolve as the trial progresses and new information about the system’s behavior emerges. “Documents must anticipate questions clinicians cannot answer and provide accessible responses within the text itself,” Su writes. The framework is presented as one of many possible approaches, but its underlying logic is pointed: consent documents for AI trials should be engineered for uncertainty rather than pretending it away.

What gives the argument its force is the deeper imperative Su identifies. “Participant autonomy in AI trials depends on information ecosystems designed for the world as it actually functions, not the world as we wish it were,” he writes. In other words, autonomy is not a box checked by a signature; it is a property of the entire information environment surrounding a trial. If a participant cannot understand what an algorithm might do to them, or why, then the signature on the consent form records agreement to something the participant never truly grasped. Su’s framework treats the consent document as a piece of designed infrastructure, subject to the same standards of usability and honesty as any other component of the trial. It is a technical argument with an ethical core, and it lands at a moment when AI is moving from the periphery of clinical research into its center.

The second feature shifts from the research lab to the physician’s office. JMIR Correspondent Wendy Glauser reports on a physician-led campaign in favor of Canada’s Bill C-34, the proposed Safe Social Media Act, in a piece titled “Canadian Medical Association Calls for Social Media Policies to ‘Heal Childhood.'” If passed, the bill would ban children under 16 from creating social media accounts unless the platform can demonstrate that it has adequate safeguards in place to protect children’s mental health. It would also require AI chatbots and social media platforms to better prevent cyberbullying and content that encourages self-harm or sexually victimizes children. Glauser’s reporting captures the mechanics of the legislation in precise terms, but the campaign’s emotional engine is the uptick in mental health issues that physicians describe observing in their own offices, an observation that has turned a policy debate into a professional cause.

The feature does not flatten the science to fit the campaign. Glauser speaks with psychology researcher Amori Mikami, PhD, about the genuine complexities of social media’s effects on mental health, effects that are not uniformly harmful and that vary across children, platforms, and patterns of use. That nuance matters, because the bill’s safeguards will have to be designed against a moving target: the platforms themselves are built on recommendation algorithms and AI chatbots whose behavior changes with every model update. The Canadian Medical Association’s support for the legislation places physicians in an unusual position, advocating not for a new drug or device but for regulatory constraints on the digital environments their young patients inhabit. The campaign’s framing, healing childhood, signals that the profession sees the problem as systemic rather than individual, a matter of policy design rather than parental vigilance alone.

The third feature moves from legislation to the bedside, and from Su’s and Glauser’s correspondents to one with direct clinical experience. Jenna Congdon, a JMIR Correspondent and ICU nurse, covers a promising closed-loop device for autonomous oxygen titration in “Could Autonomous Oxygen Titration Help Patients—And Busy Care Teams—Breathe Easier?” The clinical problem she describes is deceptively simple. Hypoxemic patients need their supplemental oxygen flow adjusted repeatedly to maintain a healthy blood oxygen level, and each adjustment is a small task that competes with a hundred others for a nurse’s attention. The device Congdon profiles uses feedback from a pulse oximeter to titrate a patient’s blood oxygen level to a set target, closing the loop between measurement and intervention without requiring a human in every cycle.

The evidence for the device comes from a recent clinical trial, and the headline finding is concrete: automated oxygenation control significantly reduced time spent in both hypoxemia and hyperoxemia, the states of low and high blood oxygen respectively. That dual benefit is technically significant because the two failure modes pull in opposite directions. Giving too little oxygen risks hypoxemic injury, while giving too much risks hyperoxemia, which carries its own harms in vulnerable patients. Manual titration, performed intermittently by busy staff, tends to overshoot and undershoot; a closed-loop system that samples continuously can hold the patient in a narrower band. For care teams, the time recovered from this repetitive task is not trivial, and Congdon’s reporting situates the device within the broader pressure on hospital workforces.

Yet Congdon is careful not to oversell the autonomy. Autonomous titration still operates within clinician-prescribed parameters, meaning the device adjusts oxygen flow only within limits that a human has set in advance. That design keeps a clinician in the loop at the level of policy even when the loop is closed at the level of execution. But it also raises a question that Congdon puts directly: who holds ultimate responsibility when things go wrong? If a device titrating oxygen within prescribed limits produces an adverse outcome, is the fault with the clinician who set the parameters, the institution that deployed the device, or the manufacturer whose algorithm failed to anticipate a particular patient’s physiology? The question echoes Su’s consent argument from the first feature, because both stories concern systems whose behavior is partly delegated to automation and whose accountability structures have not yet caught up.

About JMIR Publications News and Perspectives, the publisher describes itself as a leading open access publisher of digital health research, and the News and Perspectives section is the newest addition to its portfolio, established to bring the rigor and integrity of academic publishing to scientific journalism. The section features well-researched, expert-driven content from the Scientific News Editor, Kayleigh-Ann Clegg, PhD, and a network of specialist JMIR Publications Correspondents, with the stated aim of keeping the digital health community informed, inspired, and ahead of the curve. The three features released this week, on AI-era consent, on Canada’s Safe Social Media Act, and on autonomous oxygen titration, exemplify the section’s method: correspondents with domain expertise, whether in research, psychology, or intensive care nursing, reporting on technical developments with the precision the underlying science demands. As a technology organization with publishing at its core, JMIR Publications provides tools and resources beyond traditional publishing, supporting researchers at every step of the dissemination process, and its portfolio includes the renowned Journal of Medical Internet Research. The throughline of the week’s coverage is unmistakable. Whether the subject is a consent document, a federal bill, or a bedside device, the digital health field is being asked to design its systems for the world as it actually functions, and to decide, deliberately and in writing, where human responsibility begins and ends.

Subject of Research: Digital health news: AI-era informed consent, Canada's Safe Social Media Act, and autonomous oxygen titration

Article Title: JMIR News: Informed consent’s redesign for AI-related trials, an autonomous oxygenation device, and the physician-led campaign for Canada’s Safe Social Media Act

Article References: JMIR News: Informed consent’s redesign for AI-related trials, an autonomous oxygenation device, and the physician-led campaign for Canada’s Safe Social Media Act. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: informed consent, artificial intelligence, clinical trials, Safe Social Media Act, oxygen titration, digital health, JMIR, News, Informed, consent, redesign, AI-related