From ScienceMag:

Lilia Cervantes was a few years into her career as an internal medicine physician at a Denver hospital when she lost a patient who had become a friend. Hilda was an undocumented Mexican immigrant with kidney disease who did not have health insurance and could only get dialysis weekly at emergency rooms. She died in 2014, leaving behind two young sons. Cervantes felt the injustice of her death. “I decided I needed to learn to do research,” she says.

Cervantes later won an early-career development grant from the National Institutes of Health (NIH) that would pay enough of her salary that she could spend 75% of her time studying barriers to dialysis for Latino people. The grant also funded a small study within this group investigating whether meeting regularly with a community health worker could improve their health by spurring them to get the blood-cleaning procedure more often. The work helped Cervantes—now a professor at the University of Colorado—fall in love with clinical research.

But that grant mechanism will no longer fund trials, she learned at a May meeting of an NIH advisory board she sits on. She’s “devastated,” Cervantes says, noting that she is not speaking for that board or her university. “I’ve lost so much sleep.”

Some NIH staff are also alarmed by the move. “It cuts off a critical training pathway for clinical scientists,” a scarce part of the biomedical research workforce, says Jenna Norton, an NIH program officer who has protested many recent changes at the agency and is speaking in her personal capacity. But others argue the career development awards—NIH makes about 90 of these K grants each year—are not the best use of the agency’s budget, as they only provide enough funding for small, exploratory clinical studies that lack statistical power.

NIH this week announced the ban on clinical trial funding for K award recipients as part of an effort to simplify the grants meant to help early- and midcareer researchers who have mentors become independent investigators. Starting in July 2027, the awards can still be used to gain clinical trial experience in a study led by a mentor, but not to directly fund any clinical trials. The change is intended “to better align K awards with their primary purpose of career development,” the notice says. If a K recipient wants to run a trial, they must find other funding from NIH or sources such as foundations.

The agency is seeking input on “additional strategies” to support budding “scientists who can lead rigorous clinical trials,” the notice adds. Another notice seeks input on two options: a special type of clinical research grant that would be paired with a K award, which would still pay for the person’s salary and mentoring; or a two-stage award broken into a mentored portion, then a research grant. Both would allow new clinical investigators to conduct “studies with greater complexity, rigor, and statistical power,” the NIH press office said in response to a request for comment.

Some NIH observers remain concerned that jettisoning the K clinical trial awards is a mistake. Jeremy Berg, former director of NIH’s basic research institute (and former editor-in-chief of Science), has examined K grants made from 2010 to ’20 to see how many awardees went on to win an R01, the main type of NIH research grant. He found that about 23% later received an R01 for a clinical trial—more than twice the frequency as in the much larger group of nonclinical trial K awardees.

“If I were looking at this [program], I would be pretty close to dancing on the table that this was a huge success,” Berg says.

Berg also worries that marginalized groups and certain research topics may be disproportionately harmed. Almost two-thirds of those receiving the K clinical trial awards were women, and much of the funded research centered on studies on HIV, women, and behavioral research, according to a grant wording analysis conducted by Berg. Those areas “are all things that [President Donald Trump’s] administration is generally not very enthusiastic about,” he says.

Michael Lauer, who stepped down as NIH’s extramural research chief in early 2025, is among those who support the new NIH policy. He says the K clinical trial awards came up 10 years ago when NIH officials took steps to address what they described as too many “small, crappy clinical trials.” They changed standards for NIH trial funding in multiple ways but in the end left those particular K awards alone.

But Lauer’s glad to see them go now. Even small studies with human volunteers “should be directed by experienced people,” he says. In his field of heart disease, Laurer notes that he’s always advised early career researchers to get experience with the large, multiskilled teams that conduct major clinical studies and eventually move up to leading trials.

Cervantes responds that K awards still entail oversight by established investigators, but the type of research on getting proven treatments to patients that she conducted is done by smaller teams within real-world clinics and hospitals—not by the big clinical trial organizations Lauer has in mind. “My concern is that this [eliminating the awards] will lead to further attrition” from this part of the clinical researcher pipeline, she says.

NIH is taking comments on ways to support early-career clinical researchers until 30 September.

Read More

From ScienceMag:

When I recently opened up my word processor to start to compose a manuscript, I felt like I was learning to write again. I had to find the right words, the right style, the right voice. Every possible phrasing was crowding my mind, and the effort required to select and combine them made the task intimidating. I was frozen, overwhelmed. It had been more than a year since I had put in that kind of effort myself—because I had outsourced my writing to generative artificial intelligence (AI).

I had loved writing since I was a teenager. I grew up at the beginning of the internet, when blogs proliferated, and I spent most of my time after school writing posts on my own blog. My trivial stories about video games found a modest audience, but writing and being read brought me immense joy.

As an adult, I ended up as a radiologist—specializing in images, not words—but I still took pride in my writing style. Composing my first academic paper took dozens of hours, as I meticulously selected each word—then translated it, as I am not a native English speaker. I felt confident about the initial draft, but when I sent it to my supervisor he handed it back full of red notes, asking me to simplify, rephrase, reorganize. I realize now that he was helping me improve—but at the time, I felt frustrated I had not succeeded on my first try, and overwhelmed by the work of rewriting.

In the following months, however, I found a solution. I realized I could run my drafts through ChatGPT and any rough edges would instantly disappear. My writing process became much more efficient and my adviser was ecstatic about my progress. The red notes became rarer, and their content laudatory. Come 2025—the most successful year of my young career—and everyone was impressed by my scientific output. I could write the draft of a manuscript in hours, and reviewers commended the quality of my writing.

Except I did not author most of those texts. I came up with the idea, the experiments, the data, and the references. But the process was devoid of writing, and really, of me. Relying on AI felt like being in a tunnel, seeing nothing but the light of the flawless final submission at the end, oblivious about how I got there. I did not get to experience the pleasures of unraveling a thread of thought, encountering a challenge along the way, overcoming it with a creative twist, rephrasing, rereading, imagining the reader’s reaction. By the end of the year, I had a sense of vertigo: What if I could not write without assistance anymore?

I also began to realize AI was making me a less critical scientist. I would feed the chatbot my results and ask it to write the story I wanted to tell, and it would always oblige, never questioning my ideas. And although the papers it produced seemed good on the surface, once I started to actually read them in detail, I found they often said very little. I realized writing is a process that not only transforms ideas into text, but also allows the writer to continuously question and refine those ideas. The AI never did this.

So when I was invited to contribute an opinion piece for a journal a few months ago, I decided to close my internet browser and open my word processor again. I felt overwhelmed at first. But once I made peace with spending hours instead of minutes to generate a draft, muscle memory kicked in and I found myself having the best time of my academic life in more than a year. Once again, I experienced the bliss of finding the appropriate phrasing or the perfect figure of speech. I felt more engaged with my work, as I could retrace its authorship, from start to finish. And, to my great surprise, writing in my own voice brought an unexpected benefit: I felt proud of my pieces again. I showed them to my family and colleagues, and their reactions felt special. They were experiencing human emotions to human words.

When I showed my next manuscript to my supervisor, the red notes came back. But this time, I took them as a compliment. They were a marker of my own voice—and a reminder of the progress I could still make.

Read More

From ScienceMag:

A few years ago, I left a faculty meeting feeling reassured by how smoothly the conversation had gone. We discussed a project that would bring psychologists like me together with speech pathologists and health researchers to evaluate a rehabilitation program for stroke and other neurological conditions. Around the table, everyone used the same terminology and the discussion moved efficiently from one idea to the next. I took that fluency as evidence that we understood one another. But when the time came to translate our ideas into actual research plans, I was surprised to realize we had been talking past one another. We had mistaken shared jargon for shared understanding.

Early in my career, I’d absorbed the lesson that fluency equals comprehension. In academia, we are trained to value quick answers, polished explanations, and intellectual confidence. As a graduate student, I often felt pressure to fill the silent gap after
the professor had asked a question. Once, a classmate stopped me in the hallway and said, “You always seem to know exactly what to say.” At the time, I took it as a great compliment. I began to equate quick, confident responses with understanding itself.

Years later, that meeting with my colleagues made me question my assumptions. And I noticed the same pattern in a workshop I was teaching for psychology and exercise science students. When I posed theoretical questions, the psychology students responded quickly and confidently. But I saw several exercise science students furrowing their brows and glancing uncertainly at one another. When the discussion shifted to athletes’ real-world performance, the dynamic reversed. Exercise science students spoke fluently about muscle fatigue, progressive overload, and recovery cycles, whereas the psychology students became quieter, shifting in their seats as they tried to follow.

Each group sounded highly competent—until they had to explain their reasoning to the others. When that happened, the room filled with a silence that made me anxious. My instinct was to intervene immediately—to clarify, rephrase, provide structure, rescue the conversation before it became uncomfortable.

Midway through the workshop, I changed my approach. I asked students to explain a core concept from their field in a way anyone could understand. At first, almost no one spoke. Then came tentative attempts. Half-finished explanations. Students began to question each other as they revised their wording. One exercise science student abandoned technical language about recovery cycles and simply explained, “If the body never gets time to adapt, performance eventually drops instead of improving.” The psychology students began to connect the idea to stress, burnout, and cognitive overload. Gradually, the energy in the room changed. The students were no longer looking to me for confirmation. They were negotiating meaning together.

This experience changed how I approached my own collaborations. At a later meeting for the rehab project, I noticed our group repeatedly using the word “engagement” as though we all meant the same thing. I paused the discussion and asked each person to describe the term’s meaning within their field. For one colleague, engagement referred to clinicians’ verbal participation during training sessions. For another, it referred to patients’ adherence to their treatment plan. As an educational psychologist, I thought of engagement largely in terms of what was happening internally—whether people were genuinely motivated and actively processing what they were learning. What initially sounded like agreement was several distinct meanings hidden within a shared term. The conversation slowed once we realized this, but we ultimately designed a better research project because we finally understood where our perspectives diverged.

I teach, collaborate, and listen differently now. In my classes, I have become more comfortable allowing pauses, uncertainty, and unfinished explanations to remain in the room long enough for students to work through them and confront what they really mean. In interdisciplinary collaborations, I spend less time trying to move conversations efficiently toward agreement and more time paying attention to the awkward silences that indicate understanding has begun to break down. When a conversation slows, I no longer see it as a sign of failure. I see it as evidence that we are finally doing the harder work of understanding one another.

Read More

From ScienceMag:

As my colleague stood there crying in my office, telling me the details of how he had just been forced to retire, I was mortified. I was an assistant professor a year into my tenure, and he was “a name”—someone well known in the field. I felt bad for him. But as a noncrier, I was unsure where to look as he sobbed, or what to say to “fix” it. Like most men of my generation, I was socialized early on to stay composed and just deal with a problem; crying was not acceptable, because it would be perceived as weak and vulnerable. I was shocked that my colleague was allowing his tears to fall in front of me, and the panic and confusion I felt stayed with me long after.

The experience with the retiring faculty member was 2 decades ago. But I found myself in similar situations years later, in the context of mentoring conversations where I was the more senior person. In one, an early-career colleague asked about a possible career pivot. I offered what I thought was practical advice: She wasn’t ready yet. The conversation moved on. But 40 minutes later she became visibly upset and began to cry. As I sat there trying to resume the conversation, I felt the same flicker of discomfort and uncertainty about how to respond that I had experienced years earlier.

Around the same time, a colleague injured his hand in a door after a meeting in my office. He was in pain and visibly upset. This time, the tears did not unsettle me at all; they made sense as part of a problem I knew how to address. I checked he was OK and moved directly to the institutional safety procedures I had been trained to follow. Afterward, though, I faced an unsettling question: Why was my response so different when the tears reflected physical pain, not emotional distress?

Thinking about the two situations, I realized I needed to allow more space for emotional crying, and to be open to what it signaled. The cues that my early-career colleague was upset had been there—hesitation, small shifts in tone, moments of uncertainty. The crying simply made visible what had already been building. I began to wonder whether I had delivered my message too directly and what I might have done differently to lessen the emotional impact it had on her. And instead of seeing her crying as a problem, I could have taken it as an opportunity to better understand her as a person.

In academia we are rarely trained for moments when a colleague becomes upset in front of us. But this is no less important than knowing how to deal with cries of physical pain. Now, when someone I work with starts to cry, I know not to brush it aside or immediately move toward making the problem go away. I try to slow the conversation and allow space for the reaction. I pay closer attention to what led up to that moment, including what I said and how I said it. Only afterward do I return to the issue that prompted it.

I still feel traces of my earlier discomfort. But I no longer see it as something to eliminate, because the stakes are high during these moments. They can lead to a more open and honest exchange and signal an important shift in the conversation—from routine discussion to something more consequential. Tears can mark the points in work life where decisions matter most: whether to stay or leave, whether progress is enough, whether a path still feels viable.

I sometimes think back to that early encounter in my office. Today, I would tell my junior self that I don’t need to fix the person who is crying, or to run away from the emotional reaction. The tears should remind me that my job as a faculty member and colleague is not just about leading for better performance, but also about embracing the whole person behind the moment.

Sometimes, the most useful thing to do is not to stop the tears, but to notice what they might be telling us about a person—and about the work lives we share.

Read More

From ScienceMag:

I sat in the heavy silence of my office, waiting for an undergraduate computer science student whose academic path I was about to derail. I had noticed a glaring red flag in his final report for my class: It answered questions I hadn’t even asked. It was clear that artificial intelligence (AI) had invented details to fill the gaps. My student had presented a chatbot’s output as his own work, and I could not accept that. But the meeting that was about to take place made me realize a flat ban on AI use was not the answer.

I had worked with earlier forms of AI for almost 2 decades. After a Ph.D. in computer science, I moved to the private sector, teaching machines to catch financial fraudsters and optimize marketing expenditures. Yet, I never left the classroom, teaching two courses at a local university. I felt confident I was preparing my students for the future.

My dual career in tech and academia felt safe and rewarding. Then generative AI rewrote the playbook. Suddenly, I wasn’t just designing software to assist humans; I was watching technology threaten to bypass human input entirely. AI could now generate complex code in seconds. A machine was claiming intellectual territory I had thought was mine.

For the first time, my career path felt like an anxious race to outrun my own obsolescence. So I adapted, spending hours integrating new generative AI tools into my daily routine to summarize emails, draft documents, and automate tedious coding tasks. It wasn’t just professional upskilling; it was self-defense. But in the classroom, I maintained a strict no-AI policy.

At this crossroads, the student I intended to fail walked into my office. He didn’t try to hide that he had used AI to generate much of his assignment. Instead, he admitted his anxiety. He felt that mastering these tools was essential for his future career, yet he had no idea how—or even whether—he was allowed to use them. The more we talked, the more I saw my own doubts mirrored in him. My indignation gave way to a nagging sense of hypocrisy. I used AI daily—how could I expect my students to avoid it entirely? Outlawing AI had felt comfortable—a neat wall built to preserve a familiar order. I hid behind rigid rules because I was terrified to admit the ground beneath my academic feet had shifted.

Instead of failing the student, I gave him a chance to rewrite his report. But as the office door closed behind him, the full weight of our encounter settled on me. I realized that instead of building higher walls, I needed to dismantle them.

At my next lecture, I projected a blank screen onto the wall and invited my students to negotiate an “AI contract.” At first, they were guarded, but as I shared my own experience with AI, the classroom dynamic shifted. We stopped playing cat and mouse and became partners. Students opened up about their AI use and began to ask questions. After some debate, we drew a line separating mechanical churning from actual thinking. Automating repetitive tasks or literature searches was acceptable. Bypassing critical analysis was not: System architecture and design would remain strictly human tasks.

I have since made cocreating an AI contract a standard part of my classes. I moderate the discussion, but a consensus always emerges naturally. Although the terms differ slightly for each class, the contracts all insist that critical analysis should remain human.

The challenge of AI has prompted me to make other changes as well. Since the release of ChatGPT, student reports have grown steadily longer. I often felt I was evaluating a machine’s capacity for volume, not a student’s capacity for insight. So, I now limit written papers to two pages and base students’ grades more heavily on oral discussions of their work, which allows me to probe their thought process and challenge their decisions.

After doing these things for four semesters, I’ve seen benefits beyond reducing AI use. Defining the AI contract has become a powerful icebreaker, creating a collaborative atmosphere. The short reports force students to synthesize their knowledge, and oral defenses sharpen their argumentation skills.

Our real duty as teachers has always been challenging students to think. The task now is finding ways to engage them in the painful, essential work of critical thinking in a world where machines are willing to do it for them.

Read More