Failure in science is not limited to failed experiments or rejected papers: educational systems, funding pressures and institutional cultures can also narrow researchers’ ambitions and make failure harder to recover from. In the third episode of Failure in Science, Dr Stephen Webster of Imperial’s Good Science Project examines how universities can better recognize uncertainty, treat disproof as part of scientific progress, reduce inequalities in education and create safer environments for entrepreneurship and innovation.
When institutions fail researchers
“I’m Dr Stephen Webster, director of the Good Science Project at Imperial,” Webster says. The programme examines what researchers value in research culture, as well as what they believe needs to change.
The episode asks whether rigid educational systems can stifle imagination, whether competition for grants encourages researchers to play safe, and whether institutions themselves can fail the people working within them.
Jem Woods, a professor at Imperial College London, opened the Failure in Science conference in April by quoting Albert Einstein:
“As our circle of knowledge expands, so does the circumference of darkness surrounding it. And we have to continue to keep on thinking around and being critically reflective on that uncertainty and the parameters of that uncertainty and how we perceive what we know and what we don’t know.”
Woods says the quotation captures the central questions around failure in science: how researchers define the limits of knowledge, and how critically they reflect on what remains uncertain.
Rejection and the value of mistakes
Mary Ryan, Imperial’s vice provost for research and enterprise and the Armourers and Brasiers’ Chair for material science, argues that established academics need to speak more openly about their failures.
“When you see someone who’s clearly, you know, a successful academic,” Ryan says, it can be difficult for early-career researchers to imagine that person has experienced rejection. Yet every academic is likely to have a computer folder containing rejected proposals and papers.
Some rejected ideas were premature, some were incomplete and some were simply bad, Ryan says. Rejection does not automatically mean that an idea lacked value. What matters is not only how often researchers encounter rejection, but also how they respond to it.
Failure can also occur when an experiment succeeds in its intended purpose but overlooks something important in the surrounding process. Ryan points to the history of graphene as an example.
In the 1980s and early 1990s, many researchers working in surface science and scanning tunnelling microscopy used highly oriented pyrolytic graphite, or HOPG, as a standard substrate. To produce a clean surface, they used adhesive tape to cleave the graphite, which separates easily.
For about 10 years, “literally thousands of people were throwing away bits of graphene on their Sellotape,” Ryan says, because they were focused on the surface they wanted to study rather than on what the cleaning process had removed. The discovery of graphene using that technique was later awarded the Nobel Prize.
For Ryan, the episode illustrates how focusing on one objective can lead researchers to discard something more important. It is also a broader metaphor for life: concentrating on one thing can cause people to overlook what matters most.
Teaching uncertainty
Ryan says education must give students a firm understanding of established knowledge while also showing them that uncertainty remains at its boundaries. This is difficult because assessment systems tend to reward clear answers, while uncertainty can be uncomfortable for both educators and students.
She sees that discomfort as an opportunity to redesign education around difficult, open questions. Questions about the universe have inspired generations of researchers, and many remain unanswered. Treating education and knowledge as complete when they are not may prevent students from developing the curiosity that drives discovery.
That challenge is particularly important as artificial intelligence increasingly captures and categorizes knowledge. Students need to understand that neither people nor AI systems possess the totality of knowledge, Ryan says. There are still uncertainties and unanswered questions, and users of frontier technologies must ask what knowledge a system was trained on and where the boundaries of that knowledge lie.
“We have to, I think, bring it further into the centre, what is unknown, before we build on what we do know,” Ryan says.
Designing experiments to fail
Woods describes science as a discipline founded on experimentation, with the risk of failure at its core. He questions the distinction often made between curiosity-driven, or fundamental, science and applied science, arguing that curiosity drives experimentation, testing and the generation of knowledge in both areas.
Science continually narrows some boundaries of knowledge, Woods says, but researchers should also consider how to expand those boundaries and remain curious.
Ryan suggests beginning with the question an experiment is intended to answer. Is it asking why something happens, what happens or how it happens? The answer influences the design of the experiment.
In discovery science, researchers often begin with a hypothesis about why something happens. They should then design an experiment capable of disproving that hypothesis, Ryan says. In that sense, experiments should be designed to fail, because disproving an explanation is a way for science to progress.
Ryan believes this aspect of the scientific method has been weakened because such experiments can be difficult to conduct, explain and publish. Scientific systems also need better ways to record experiments that disprove hypotheses, since those results contribute to knowledge by narrowing the possible answers and identifying the most promising paths forward.
Researchers may also need to establish the limits of a method or idea before they can understand what works. Reaching the point at which something no longer works can reveal the boundary of discovery or applied science, even if “failure” is not always the most precise description of that process.
Individual and systemic failure in education
Roseanne, who has been at Imperial for around four years, says the conference session on failure in education was especially relevant because she had experienced the issues under discussion. She was interested in the tension between individual and systemic failure: systems can produce individual failures, while individual failures can also symbolize shortcomings in the system.
Students, she says, need to understand that failure carries considerable weight and symbolism in daily life, but that experiencing failure does not necessarily mean they themselves have failed.
Roseanne argues that education is underpinned by inequalities. Students do not all arrive with the same science capital or social capital, and therefore do not begin with equal access to the knowledge and confidence needed to succeed. Educational institutions need to bridge those gaps while helping students develop the confidence to make mistakes, learn from them and improve.
Ryan says Roseanne’s comments highlighted that the impact and likelihood of failure are not equally distributed. Social capital, personal background and disparities in educational inputs can affect both the likelihood of failure and the effect that perceived failure has on educational outcomes and later life chances.
“Some people are allowed to fail, and still succeed,” Ryan says. Educational institutions therefore have a responsibility to level the playing field, ensuring that talented people from all backgrounds have an equal chance of success.
One way to do that is to build resilience. A failed exam should not be treated as a judgment on a student’s talent, knowledge or overall potential. Instead, it can be a point at which the student reflects, learns and tries to improve.
Ryan acknowledges that educational systems are often built around distinctions such as pass, fail and degree classifications including a first or a 2:1. She remains committed to educating people to the highest level they can reach, but says institutions must also communicate that every student’s contribution has value and may make a difference.
Redefining success for early-career researchers
Ryan says universities also need to challenge the assumption that the ultimate destination for every researcher is a professorship. The likelihood of an early-career researcher becoming an academic professor is “probably about 5%,” she says, and professorships are not necessarily the most impactful jobs.
Imperial has established a new research institute for early-career researchers to help them consider a range of possible paths, including careers in industry, policy, publishing and the public sector. The aim is for researchers to identify these options early, not regard them as second choices after an academic career does not work out.
Ryan describes this as a reset of expectations: researchers can decide from the outset what they want to do with their skills and research, while the institution can offer a broader and more realistic account of success.
A safe space for entrepreneurial failure
Innovation and entrepreneurship are inherently high-risk activities. Ryan says they should involve risk, including the possibility of “failing fast”, but academic culture can make students reluctant to fail because they are accustomed to being assessed through exams and tests.
Imperial’s undergraduate entrepreneurship programme is run through the Enterprise Lab. It helps students develop ideas into devices or companies, connects them with entrepreneurs outside the university, provides mentors and runs programmes and competitions.
The activities are deliberately extracurricular and do not count towards academic credit. Students can explore ideas without being graded, whether those ideas are unsuccessful or develop into companies. They can decide whether entrepreneurship is right for them without receiving an academic mark for the outcome.
That changes the culture around experimentation. The programme provides a safe space in which failure can be a positive result, because an unsuccessful idea may still teach students something valuable or reveal that they want to pursue a different direction.
“Have an idea, try something, fail fast. No consequences other than you learnt something,” Ryan says. She regards that approach as an important part of entrepreneurial education.
Changing the culture around failed start-ups
Ryan says the same questions apply to university spinouts and the wider innovation system. The San Francisco Bay Area is often described as a centre of innovation, partly because entrepreneurs can create several start-ups over time. A failed start-up does not necessarily prevent them from securing funding for another venture.
When Ryan speaks to entrepreneurs in the Bay Area, she says, they often begin by describing the three companies they had previously started that failed. In the United Kingdom, and probably Europe more generally, investors are more likely to demand a track record. In practice, that can mean a record of successful ventures rather than experience that includes failure.
The difference creates two contrasting systems. In one, an entrepreneur may feel that a single opportunity must work. In the other, they can try three or four times before one succeeds.
Ryan says that if other regions are to catch up with the United States, the culture needs to change. That may require more venture capital and investors willing to take greater risks, as well as greater resilience among individuals. When one idea does not work, entrepreneurs should be able to learn from it and try again.
For Webster, Woods, Ryan and Roseanne, accepting failure is not about lowering standards. It is about recognizing that uncertainty, rejection, mistakes and unsuccessful experiments are inseparable from learning and discovery—and that the systems surrounding researchers can determine whether those experiences become setbacks or opportunities to progress.
Source: www.nature.com