Not how science got lost because scientists are corrupt. Not how science got lost because models are useless. Not how science got lost because climate change is not real. Those are not the stories I am interested in.

The story I am interested in is this: how does science get lost when most of the people involved believe they are doing the right thing?

Let’s start this story from the beginning. RCP8.5 began as a high-emissions climate scenario. It was useful for asking what the world might look like under a severe emissions pathway. As a stress test, it had value. As a warning sign, it had value. But over time, it was turned into “business as usual”, representing the likely future absent dramatic policy action. That is the point Zeke Hausfather and Glen Peters made in Nature, arguing that the worst-case scenario should no longer serve as the default baseline, because better policy requires more realistic assumptions.

That distinction matters. A worst-case scenario is not the same as a baseline. A stress test is not the same as a forecast. A scientifically legitimate tool can become misleading when it moves from one context to another without the appropriate caveats.

That was 6 years ago, 3 years after Ritchie and Dowlatabadi showed that the underlying assumptions were no longer plausible.

That is where the RCP8.5 story becomes bigger than RCP8.5.

The Intergovernmental Panel on Climate Change was created in 1988 by the World Meteorological Organization and the United Nations Environment Program and has since become the central global body for assessing climate science for governments. But the IPCC was never simply a scientific academy. It was political by design, not partisan but intergovernmental. Its authority rests precisely on the hybrid arrangement between science and governments. That arrangement gives its assessments unusual power. It also makes boundary maintenance difficult.

The United States played a major role in creating and shaping this system. The U.S. came first before the UN, not the other way around. That history matters because it reminds us that climate science assessment was not born outside politics and later corrupted by it. It was built from the start as a mechanism for translating scientific knowledge into governmental judgment, and the U.S. was in the lead. That may have been necessary. It may even have been wise. But it also means that the relationship between climate science and climate politics has always required careful boundary management.

The problem is that, over time, managing that boundary became harder.

In the United States, science itself has become increasingly sorted by party. Dan Sarewitz warned more than fifteen years ago that the overwhelming Democratic identification among scientists was a problem because a scientific community so closely aligned with one political coalition becomes easier for others to distrust. More recent evidence points in the same direction: political donations by U.S. scientists have become overwhelmingly Democratic, with less than 10 percent of scientists’ federal campaign donations going to Republicans in recent years. That does not prove that science is wrong. It does suggest that the social location of science has changed.

Public trust data tell a similar story. The problem is not simply that Americans stopped trusting science. Trust remains fairly high overall. Pew reported in 2024 that 76 percent of Americans had a great deal or fair amount of confidence in scientists to act in the public interest. But the partisan gap is now large: 88 percent of Democrats expressed confidence in scientists, compared with 66 percent of Republicans.

That is the crucial shift. The issue is not merely declining trust. It is a redistribution of trust. Science has not disappeared as a source of authority; it has become more politically legible. One side increasingly views science as part of its coalition, while the other increasingly views it as part of the opposing coalition.

Climate science sits at the center of this tension.

This is where I think the RCP8.5 controversy becomes especially important. It offers a concrete example of how a scientifically useful scenario can become politically useful in ways that make correction harder. A high-end scenario dramatizes risk, creates urgency for action, generates headlines, and supports a narrative that things are worse than expected, that time is running out, that compromise is dangerous, and that hesitation is complicity.

But what happens when the most politically useful scenario is no longer the most scientifically plausible baseline?

That is the uncomfortable question.

The answer, I suspect, is that correction becomes morally and politically costly. Calling RCP8.5 no longer “business as usual” could be seen as giving ground to climate skeptics. Saying that some emissions trends have improved relative to older worst-case assumptions could be treated as weakening the case for action. Calling for more realistic baselines could be mistaken for complacency.

So, the correction comes late. It comes after the scenario has already circulated through thousands of papers, news stories, risk assessments, advocacy campaigns, corporate disclosures, policy arguments, and campaign promises. Science eventually self-corrects, as it did in this case, but only after political usefulness has raised the cost of correction.

That may be the real lesson: science does not stop being self-correcting all at once. It becomes slower to self-correct in areas where correction is interpreted as betrayal.

This pattern is not limited to RCP8.5. Climate politics has repeatedly struggled to distinguish between evaluating options and morally disqualifying them in advance. Adaptation was once treated by some advocates as a dangerous distraction from mitigation, a moral hazard. It then became unavoidable. Nuclear energy was long resisted by many environmentalists, even though it is a low-carbon energy source. It is now being reconsidered in many circles of climate policy. Geoengineering, especially solar geoengineering, occupies the current version of this position: not simply an object of scientific assessment, but often a test of political loyalty.

None of this means adaptation is sufficient, nuclear is easy, or geoengineering is wise. Each raises hard questions. But that is precisely the point. The job of assessment is to ask those questions carefully, not to decide in advance which questions are morally admissible.

When science-policy communities treat some options as unspeakable, they make science less useful.

Roger Pielke Jr.’s experience is relevant here. He argued for climate action long before the activist community accepted adaptation. But he also challenged claims about disasters, damages, and the attribution of extreme events when he thought the evidence did not support the rhetoric. For that, he became, in his own account, a climate heretic. Whether one agrees with him on every point is not the issue. The issue is whether a healthy science-policy system can tolerate internal criticism without treating it as defection from the party of science.

A mature scientific culture should be able to say two things at once: climate change is real and serious; some claims made in the name of climate urgency are overstated, premature, or poorly supported.

If it cannot say both, it has a problem.

The Biden administration’s whole-of-government climate agenda may be another expression of the same difficulty. The problem was not the political aspiration. But combining climate policy, industrial policy, infrastructure deployment, and environmental justice into a single whole-of-government framework created an enormous implementation burden. It required federal agencies to move money, build projects, measure benefits, secure local legitimacy, and satisfy multiple political constituencies at once.

That is a difficult governance problem under the best conditions. Under polarized conditions, it becomes fragile. If projects are slow, if communities resist, if benefits are hard to measure, or if implementation becomes procedurally overloaded, the political coalition that created the program may have little to show by the time electoral politics change. Then programs that took years to assemble can be reversed quickly.

That, too, is a lesson about science getting lost, not because the science of climate change is wrong, but because the politics attached to it becomes overengineered, under-deliberated, and institutionally vulnerable.

So where does that leave the IPCC and RCP8.5?

I am tempted to say that the science-policy ecosystem helped create the conditions that allowed RCP8.5 to remain in use for too long. Not because anyone issued an order. Not because there was a conspiracy. But because the incentives aligned.

Researchers could publish high-impact risk studies. Journalists could tell urgent stories. Advocates could mobilize the public. Democratic politicians could justify ambitious policy. Agencies could plan for severe futures. And critics could be dismissed as undermining the case for action.

In that environment, the scenario did not need to be imposed. It only needed to remain in circulation.

That may be the subtler and more important failure.

Science gets lost when its tools are misused. It gets lost when caveats are omitted because they weaken the message. It gets lost when uncertainty is treated as a communications obstacle rather than a condition of honest assessment. It gets lost when politically convenient claims are defended longer than they should be because correcting them would help the “wrong” people. It gets lost when criticism from within the broad climate-action camp is treated as indistinguishable from denial.

And science gets especially lost when it forgets that public trust does not come from always sounding certain. It comes from being willing to correct oneself visibly.

The RCP8.5 debate should not be used to claim that climate change is exaggerated or that climate policy is unnecessary. That would be the wrong lesson. The better lesson is more demanding: climate action needs stronger science-policy boundary management, not looser standards in the name of urgency.

A democratic society needs climate science that informs action without blurring into climate politics. It needs scenarios that clarify choices, not those that quietly become slogans. It needs assessments that preserve the distinction among plausible futures, high-end risks, and mobilizing narratives. It needs scientists, policymakers, advocates, and journalists who can say: this risk is real, but this particular claim was overstated; this policy goal is worthy, but this implementation design is weak; this criticism is inconvenient, but not necessarily wrong.

That is not a retreat from climate action. It is a prerequisite for durable climate action.

The uncomfortable possibility raised by RCP8.5 is that the greatest threat to science in politics is not always an external attack. Sometimes it is internal over-identification: science becomes so attached to a political project that it loses the ability to discipline itself.

When that happens, science does not merely become politicized. It becomes vulnerable, making it an easy target for its opponents. It turns correction into a scandal and makes trust harder to rebuild, especially in an information environment without honest knowledge brokers.

The question, then, is not whether science and politics can ever be separated. They cannot, not fully. The question is whether we can rebuild the boundary practices that keep science useful to politics without making it captive to it.

RCP8.5 is a warning. Not the warning it was often used to deliver, but perhaps a deeper one: even good science can be lost when its political usefulness outpaces its epistemic care at the boundaries of science, policy, and politics.

Three times, in three different settings, I found university presidents not just staying but listening. Not ceremonially, but substantively. Not as figureheads, but as participants in conversations that were, at times, uncomfortable, contested, and unresolved.

And each time, it caught me off guard.

The most recent instance was at Georgia Tech’s inaugural symposium on technology and civic leadership. The morning featured a high-profile dialogue between Cornel West and Robert George, an expansion of the arguments in their most recent book, Truth Matters: A Dialogue on Fruitful Disagreement in an Age of Division. When we broke for lunch, the hierarchy of the day had been established: the host, the President, the Chancellor, and then the main attraction, Robert George and Colonel West, all had taken their turns on the podium.

I was scheduled to speak on the second post-lunch panel, “Informing Public Debate: Lessons from the Intersection of Technology and Policy.” Sitting in the fifth row, listening to the panel ahead of me, I noticed something unusual: West and George were still seated in the front row, listening and engaging. They were modeling what they just preached about intellectual humility, that we are “listening in a truth-seeking spirit.” To me it meant actively and attentively listening to others, not only those with whom we may disagree, but also those who may not be our professional or intellectual equals, who may come from other disciplines or knowledge traditions, those who may not dwell in the high-minded ideal but the messy and mundane practical, in this case not just the why and what of civil discourse, but the where and how, the tasks left to the “implementers”, the engaged academic outliers like my fellow panelists and me.

I took note of that, determined to mention it if the two luminaries were still there when our panel’s turn came.

But what I did not anticipate was what I saw when we were seated to begin our panel.

In the front row but not in the center, to my left, sat Georgia Tech President Ángel Cabrera—alone in a depleted row of reserved seats, attentive, taking notes. No entourage. No signaling. Just presence.

I wasn’t entirely certain it was him, since I had just seen him for the first time that morning. I decided not to call him out directly when I mentioned George and West. Instead, I made a broader point, building on his concluding line: an addendum to President Lincoln’s Gettysburg Address; we needed a government not just of, by, and for the people, but also with the people.

He was doing exactly that. The chief executive was “with” the implementers, listening, and practicing intellectual humility.

Beyond the Podium

As I mentioned in my lead, this was not my first recent encounter with university presidents in the audience.

The next most recent instance was last November, at Johns Hopkins Bloomberg School of Public Health, a JHU-AEI symposium on science, policy, and public trust in the wake of COVID.

President Ron Daniels opened with remarks. That was expected. What followed was not.

He stayed through panels that moved from generalities to contested terrain, including questions of trust, failure, and institutional credibility. When the formal program ended, he didn’t disappear. He joined the speakers and organizers for dinner, extending the conversation.

In a forum focused on public trust, his presence did what no speech could. It suggested that engagement was not performative. It was sustained.

Listening Before Leading

The third instance occurred at a National Academies workshop on Building Institutional Capacity for Engaged Research, in June, 2024. Here, the sequence was reversed. Panels first, reflections later.

Penn State President Neeli Bendapudi listened through hours of discussion about tensions among institutional incentives, community engagement, and what counts as knowledge before offering closing reflections.

What stood out was not only her remarks but also the listening that preceded them. During breaks, she engaged directly, asking questions and following up. At one point, she asked me to share my remarks on the long-standing tension between knowledge production and knowledge use in American universities.

It was a small exchange. But it reflected something larger: a willingness to engage before reflecting.

Looking Back at What They Said

These three moments stayed with me, not simply because of who these leaders were, but because of what they chose to do with their time and attention. In a sense, they had given me and others in those rooms the honor of being heard.

This is not the norm. The public compares universities to the ivory tower, but there are ivory towers within the ivory tower. I worked under a university president who wouldn’t meet privately with anyone who didn’t have “President” or “CEO” in their title.

So, I wanted to return the honor these three exemplars had just extended to my contemporaries. I went back and revisited their remarks. This time, not as a participant moving through a packed agenda, but as an observer trying to understand what linked them. What made these leaders stand out in a crowded landscape of capable, accomplished university presidents? What connected their presence in the audience to the ideas they expressed from the podium?

Cabrera: Civic Courage in a Technological Age

President Cabrera’s remarks at Georgia Tech make clear that this kind of presence is not incidental; it is also philosophical. He framed the university not as a purely technical institution but as a civic one, rooted in the unfinished democratic experiment. Drawing on Frederick Douglass and Martin Luther King Jr., he argued that democracy requires not retreat but engagement, “open dialogue, honest argument, genuine listening,” even amid disagreement.

What is striking is how directly this connects to our technological moment. Cabrera’s concern is not only polarization but also how digital infrastructures amplify division when optimized for outrage rather than understanding. In that context, the university’s role is not to withdraw into expertise but to actively cultivate the conditions for democratic discourse. His presence in the audience, listening and taking notes, was not symbolic. It was an enactment of that civic commitment.

Daniels: Universities as Arenas for Democratic Contestation

Ron Daniels’s remarks sharpen this argument by placing universities at the center of democratic life, precisely because of the current crisis of confidence. He openly acknowledged that confidence in scientific expertise has eroded and that universities have not fully met their responsibilities. But his response is not defensiveness or insulation; it is openness. Universities, he argues, must be places where even uncomfortable and dissenting perspectives receive a fair hearing and where ideas are tested across lines of difference.

His emphasis on collaboration across ideological divides reflects a deliberate effort to resist intellectual siloing. This is not about consensus; it is about legitimacy. In a pluralist democracy, trust is not restored by asserting authority, but by demonstrating a willingness to engage criticism and competing values. His decision to remain throughout the event mirrored that commitment.

Bendapudi: Rebuilding Trust Through Public Engagement

Neeli Bendapudi’s intervention addresses the same problem from an institutional and strategic perspective. She identifies declining trust in expertise and skepticism toward higher education as an existential challenge not only for universities but for democracy itself. Her response is to advance public impact research while insisting that expertise alone is insufficient. Universities must be seen as both competent and well-intentioned.

This reframes the university’s role from knowledge production alone to building relationships with society. Engagement is not an add-on; it is central to restoring legitimacy. Her emphasis on systemic solutions and institutional change reflects recognition that this work must be sustained, not episodic. Her listening to and engaging with the participants throughout the workshop were consistent with that vision.

A Common Thread: Leaving the Ivory Tower

Taken together, these three perspectives converge on a common insight: universities cannot retreat from society at a moment when society is questioning them. The response to anti-intellectualism, populist skepticism, and declining trust is not to defend the ivory tower but to leave it.

What I witnessed in these three instances was a form of leadership that accepts the risks of engagement, entering spaces of disagreement, exposing oneself to critique, and relinquishing the comfort of institutional distance.

Cabrera frames this as civic courage in a technologically fragmented public sphere. Daniels frames it as the university’s obligation to host and sustain democratic contestation. Bendapudi frames it as rebuilding trust through public engagement and institutional transformation.

All three points in the same direction.

In a democracy under strain, the role of the university is not only to produce knowledge but to create, as Robert George insisted, first the conditions under which “settled” knowledge can be questioned, debated, verified, and if necessary, unsettled, and second to commit

most earnestly to virtues like intellectual humility. That requires more than speaking.

It requires showing up. Listening. And staying in the room when the conversation becomes difficult, or different, or not on the same level. It requires leading by example.

That idea built the postwar research system and shaped institutions like the National Science Foundation. But it also created a misleading story about how American science actually works.

Vannevar Bush won the battle. Harley Kilgore won the war.

Kilgore, the West Virginia senator who conflicted with Bush over the design of the postwar research system, believed that publicly funded science should more directly serve public aims, such as regional development, industrial innovation, and widespread distribution of benefits. His institutional proposals were rejected in the 1940s. However, the underlying rationale kept resurfacing, gradually influencing American science policy.

The reason becomes clear when you step back from the post-1945 perspective. The tension between producing knowledge and using it did not start with Bush and Kilgore. It runs throughout the entire history of the American republic. The Constitution itself ties intellectual property to the “progress of science and useful arts.” Jefferson sent Lewis and Clark west not just to explore, but to gather information for governance and economic growth. The Morrill Act of 1862 created land-grant universities focused on agriculture and the mechanical arts. “Science for useful purposes” was not a late addition to American science. It was present from the beginning.

What changed after World War II was not the disappearance of this tension but the way it was managed.

Over time, Kilgore’s logic was reintroduced into the system through a series of reforms. The Bayh–Dole Act of 1980 encouraged universities to commercialize federally funded discoveries in response to economic stagnation and international competition. Engineering Research Centers in the 1980s aimed to foster closer collaboration between universities and industry. The CHIPS and Science Act of 2022 once again increased federal science funding but largely framed it around technological competitiveness and national security.

None of these reforms replaced the earlier system. They accumulated on top of it.

The outcome is a path-dependent scientific enterprise that aims to advance discovery, promote commercialization, enhance national security, support regional development, and maintain global technological leadership. Each goal is valid. Collectively, they create a system that is powerful but increasingly challenging to govern coherently.

At the same time, the nature of science itself has changed.

Bush’s model assumed a world of small research groups led by individual investigators. Today’s discoveries come from large collaborations involving universities, corporations, and governments across continents. The CRISPR patent battles—linking UC Berkeley, the Broad Institute, MIT, Harvard, and international partners—show how scientific discovery now happens within a complex ecosystem of intellectual property law, venture capital, and geopolitical rivalry.

As science grew, so did its administrative structures. Each attempt to make science more accessible to society—through commercialization, accountability standards, technology transfer offices, security reviews, and regional innovation initiatives—generated new middlemen. Program managers, compliance offices, and technology-transfer bureaucracies multiplied.

What began as a wall separating science from society gradually became a system of doors.

And every door required a gatekeeper.

Today, those gatekeepers have become the political targets. Critics increasingly depict the scientific establishment as an insulated elite system—culturally narrow, politically aligned, and structurally exclusionary. The populist backlash that has followed is less about democratizing science than about struggling to control its institutional chokepoints. Universities and research agencies now find themselves caught between competing claims about who truly represents “the public.”

Under these conditions, returning to the status quo isn’t feasible. Nor is marginal reform likely to be successful. Rebuilding the same structures or simply replacing the gatekeepers will only recreate the same tensions.

To envision a different direction, it helps to look back at a technological debate from ten years ago.

Around 2017, autonomous vehicles were widely expected to replace human drivers entirely. Industry forecasts promised a transformation similar to the shift from horse-drawn transport to the internal combustion engine. However, participatory technology assessment exercises conducted at that time revealed something striking: many members of the public preferred conditional automation, what engineers call Level-3 autonomy. They did not want to give up control completely. Instead, they wanted systems that assisted human judgment while still allowing them to intervene.

Looking back, that preference seems surprisingly foresighted. Fully autonomous vehicles are still out of reach, and transportation systems are shifting toward hybrid setups that blend automation with human supervision.

The future of artificial intelligence may look similar.

AI will not simply replace humans, nor will it leave social institutions untouched. Instead, it will reshape work, governance, and everyday life through a long transition in which human judgment and machine capability coexist uneasily.

Our future may be Level 3.

The challenge for democratic societies won’t be stopping technological change but creating institutions that can manage it. That means building infrastructure for public involvement with science and technology on a large scale, such as participatory science, participatory technology assessment, and processes that let citizens discuss technological futures alongside experts.

There are already signs of what this could look like. In 2014, NASA used participatory technology assessment to involve citizens in discussions about whether the agency should pursue a mission to redirect a near-Earth asteroid. Participants from two locations in the United States debated the scientific, economic, and ethical implications of planetary defense. The exercise did not replace expert judgment. Instead, it complemented it, showing how citizens consider risks, benefits, and public values when faced with frontier technologies.

That type of engagement may become crucial as societies face the governance challenges brought by artificial intelligence, biotechnology, and climate intervention technologies.

If reform is to start anywhere, it might not be in Washington, Boston, or Silicon Valley, the ecosystems that is most deeply connected to the current system. Genuine institutional experimentation could arise at the state level, where science policy can more directly link with regional economies and communities.

Two and a half centuries of American history suggest that the tension between knowledge and use will never disappear.

Bush won the battle. Kilgore won the war.

The next settlement between science and democracy will depend on whether we can build institutions capable of navigating a Level-3 world where expertise and democratic engagement evolve together.

The endless frontier was discovery.

The next frontier is participation.