Wang, Baranasi. 2021. The science of science.
Книга по науковедению. Большинство выводов обосновано данными статистики - по цитированию, возрасту и полу авторов и т.п.
Это новое науковедение, у истоков - Кун. Скорее социология науки. Базовую точку зрения авторов можно примерно передать так: есть множество случайных людей с разными случайно распределенными способностями, есть множество случайно распределенных научных тем, среди которых есть "выигрышные билеты", будущие "открытия". И эти грибники случайным образом ищут на тематических полях свой успех.
Эта модель оказывается во многом похожей на правду. Уже это позволяет понять, насколько современная "наука" не похожа на некоторые идеальные образы прошлого. Можно видеть, как на протяжении ХХ в. менялась эта область культуры, как наука переставала быть тем, что понимали прежде под этим словом, и становилась чем-то иным.
Там в середине очень занятные картинки. Говорится, что больше успеха приносят работы с множеством соавторов (до 1000). Большие коллективы делают сейчас науку, у них больше шансов на успех, так делают всё больше статей и т.п. И рядом - шанс статьи "разрушить" науку, то есть прервать инерционную традицию и совершить некий прорыв. Оказывается - вот странно - что такой шанс выше у статей с малым числом соавторов. Прямо парадокс.
В общем, позволяет составить представление о изменениях в науке примерно с середины ХХ в. до 2007 г., как наука переформатировалась и становилась тем, чем она сейчас является и для чего, в общем, нет ни слова, ни смысла.
There are two plausible explanations for this shift. The f i rst hypothesis is that the life cycle of innovation has changed, so that great minds now innovate at a later stage of their career. This could be due to the extended length of time required for education, which delays the onset of active innovative careers. The second hypothesis reasons that the upward age trend simply ref l ects general demographic shifts. In other words, if everyone in the world is getting older on average, science as a profession should be no exception.
To understand the origin of the observed shift, Jones proposed the “burden of knowledge” theory [88, 104, 105]. First, innovators must undertake suff i cient education to reach the frontier of knowledge.
Second, because science has been growing exponentially, the amount of knowledge one has to master to reach that frontier increases with time.
This theory offers a fresh perspective of Newton’s famous remark about “standing on the shoulders of giants”: In order to stand on a giant’s shoulders, one must f i rst climb up his back. But the greater the body of knowledge, the longer the climb.
Widely held is the idea that younger researchers are more likely to make radical departures from convention. Max Planck vividly put it this way:
“A new scientif i c truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die, and a new generation grows up that is familiar with it.” This school of thought argues age as an important factor driving scientif i c progress. In other words, if young and old scientists have different aff i nities for accepting new ideas, then scientif i c progress will have to wait for older scientists to fade in relevance, before the discipline can move forward. Yet by contrast, another school of thought argues that new ideas triumph when they are supported by more empirical evidence than the competing hypotheses. Therefore in science, reason, argument, and evidence are all that matter, suggesting that the age factors may not be as important as we thought. So is what Planck said true for science?
Although people have long suspected that younger researchers are more willing to make radical departures from convention, there has trad-itionally been limited evidence supporting Planck’s principle. For example, Hull and colleagues found that older scientists were just as quick to accept Darwin’s theory of evolution as younger scientists [114]. This f i nding supports the idea that scientif i c progress is self-correcting, guided only by truth. Yet a study focusing on 452 prominent life scientists who passed away suddenly while at their peak of their career, offers empirical support for Planck’s principle [115]. After the unexpected death of a prominent scientist, her frequent collaborators – the junior researchers who coauthored papers with her – experience a sudden drop in productivity. At the same time, there is a marked increase in published work by newcomers to the f i eld, and these contributions are disproportionately likely to be highly cited.
They are also more likely to be authored by young outsiders to the f i eld.
These results are also consistent with a study that investigates the link between age and the adoption of new ideas. By measuring new words used in a paper as a proxy for new ideas, researchers found that young research-ers are much more likely than older scientists to tackle exciting, innovative topics – and when they do, the resulting work has a higher impact [111].
These results document a “Goliath’s shadow” effect in science: Outsiders are reluctant to challenge the leading thinker within a f i eld while he is still alive, but stream in and take over leadership once he has passed away.
The fact that the two distributions in Fig. 5.2 are the same indicates that variations in P(t*) are fully explained by changes in productivity throughout a career. Indeed, the randomized curve meas-ures the variations in productivity during a scientist’s career. It shows that in this sample of scientists, productivity has a peak at year 15 of a career, and it drops rapidly after year 20. This means that young scientists have a disproportionate number of breakthroughs early in their career not because youth and creativity are intertwined, but simply because they’re in their most productive period. In other words, when we adjust for productivity, high impact work will occur randomly over the course of a career. We call this the random impact rule
Staying with the card analogy, imagine you draw one card at a time from your deck, but with varying frequency. You start furiously in the beginning of your career, drawing one card after another, rapid-f i re, excited by your work and hoping to f i nd that ace. This rapid-f i re period is then followed by a gradual decline, where you slow down how frequently you reach out to the deck. Now, if the deck is well shuff l ed beforehand, and you draw a lot more cards during the f i rst 20 years than during your later period, when will you most likely encounter the ace? During the f i rst 20 years, of course. In other words, the f i rst two decades of your career are not more creative than the later 20 years.
You draw an ace early in your career simply because you try harder.
To what degree do teams contribute to the production of science in the twenty-f i rst century? The answer is provided by a study that explored the authorship of 19.9 million research articles and 2.1 million patents [136], revealing a nearly universal shift towards teams in all branches of science (Fig. 8.1a). For example, in 1955, nearly half of all science and engineering publications were by single authors, but by 2000, the number of solo-authored papers had dwindled dramatically, while team-authored papers now made up 80 percent of all publications.
Importantly, the shift toward teams is not simply driven by the fact that the experimental challenges are becoming larger, more complex, and more expensive. Pencil-and-paper disciplines like mathematics and the social sciences exhibit the same patterns. Teams wrote only 17.5 percent of social science papers in 1955, but, by 2000, team-based papers became the majority, reaching 51.5 percent – witnessing the same trend toward teamwork as had been observed decades earlier in the natural sciences.
But perhaps more interesting than the trend itself is the kind of research that team work has produced. Teams do not merely produce more science; they are increasingly responsible for discoveries with larger impacts [136]. Indeed, on average, team-authored papers garner more citations than single-authored work at all points in time and across all broad research areas (Fig. 8.1b).
The universal shift toward larger and larger teams across science and technology suggests that large teams, which bring more brain-power and diverse perspectives, will be the engines for tomorrow’s largest advances. Research has consistently shown that as teams become larger, their products – be they scientif i c papers or patented inventions – are associated with higher citation counts [136, 197]. These trends seem to offer a simple prescription for the future of science: bigger is always better.
Yet there are reasons to believe that large teams are not optimal for all tasks. For example, large teams are more likely to have coordin-ation and communication issues – getting everyone onboard to try out an unconventional hypothesis or method, or convincing hundreds of free-thinking individuals to change direction at once, is often challen-ging. Psychology research shows that individuals in large groups think and act differently. They generate fewer ideas [198, 199], recall less learned information [200], reject external perspectives more often [201], and tend to neutralize one another’s viewpoints [202]. Large teams can also be risk-averse, since they have to produce a continuous stream of success to “pay the bills” [203].
All of which raises the question of whether relying on large teams is truly a one-size-f i ts-all strategy for producing groundbreaking science. Indeed, new evidence suggests that team size fundamentally dictates the nature of work a team is capable of producing, and that smaller team size confers certain critical benef i ts that large teams don’t enjoy.
Our analysis shows that over the past 60 years, larger teams have produced research articles, patents, and software products that garner higher impact than smaller teams. Yet, interestingly, their dis-ruptiveness dramatically and monotonically declines with each add-itional team member (Fig. 12.4). Specif i cally, as teams grow from 1 to 50 members, the disruptive nature of their papers, patents, and products drops by 70, 30, and 50 percentiles, respectively. These results indicate that large teams are better at further developing existing science and technology, while small teams disrupt science by suggesting new prob-lems and opening up novel opportunities.
But is the observed difference in the work produced by large and small teams really due to team size? Or can it be attributed to differences in other confounding factors? For example, perhaps small teams generate more theoretical innovations, which tend to be more disruptive, and large teams generate more empirical analyses, which are more likely to be developmental. Or, maybe there are differences in the topics that small and large teams tend to tackle. Another possibility: Perhaps certain types of people are more likely to work for smaller or larger teams, thus changing the outcomes associated with each.
Do large and small teams turn to different sources when con-ducting research? To answer this question, we measured how deeply small and large teams build on past literature by calculating the average age of references cited. We f i nd that solo and small teams were much more likely to build on older, less popular ideas. This is likely not a function of knowledge: Since larger teams have more members, their expertise spans a broad range of subjects; as such, their members were probably just as aware of older, less known work as the scientists who work within small teams. However, large-team scientists tend to source their ideas from more recent, and higher-impact, work. Consequently, large teams receive citations quickly, as their work is immediately relevant to contemporaries. By contrast, smaller teams experience a much longer citation delay, but their work tends to persist further into the future, achieving a more enduring legacy.
Second, the practice of listing authors based on their contribu-tions is not always exercised [51]. Sociology and psychology, for example, do not follow the tradition of putting the most senior author last. In these disciplines, being the last author really is the least desirable position. Moreover, it is not uncommon for a mentor to be in the f i rst-author slot for the crucial guidance, intellectual contributions, and f i nancial leadership of the project, even if the trainee may have done all the heavy lifting.
The trend of growing author lists has a troubling side effect: the rise of “guest authors,” those who made only very minimal contributions to the research.
A recent survey of 2,300 lead authors found that a shocking 33 percent of scholarly papers in the biological, physical or social sciences had at least one author whose contribution did not meet accepted def i nitions for coauthor-ship [166, 222]. The incidence of “undeserved” coauthors increases from 9 percent on papers with three authors to 30 percent on papers with more than six authors [213], with the most commonly cited reason for accepting undeserved authorship being academic promotion [141].
A far more serious concern are “ghost authors,” individuals who made substantive contributions to a publication but who were not acknowledged as authors [223]. Take, for example, Robert Boyle, the most acclaimed chemist in seventeenth-century London. His laboratory was populated with assistants who tended distillations, amalgamations, and rectif i cations, made observations, and recorded them for Boyle’s use [224]. Yet we know absolutely nothing about these researchers or their contributions to the research – not even their names – since Boyle pub-lished all their f i ndings and contributions as his own (Fig. 13.3).
Unfortunately, this phenomenon has not abated in the centuries since. A recent estimate indicates that more than half of the papers in many disciplines have at least one ghost author [222]. The primary victims are often graduate students, whose relegation to ghost author status can hurt their chances of career advancement.
As scientists look back upon the twentieth century, we marvel at the discoveries and inventions that our predecessors have made, from internal combustion engines to computers to antibiotics. The tremen-dous progress in these areas could imply diminishing returns, as the sweetest, juiciest apples are already harvested. Yet this does not seem to be the case. The data suggests that we are poised to discover and invent even more in the next 20 years than we have in all of scientif i c history.
In other words, even after a century of breakneck progress, science today is fresher and more energetic than ever.
How can science continue to run tirelessly after a century of exponential growth? Because unlike a car or a colony of bacteria, science runs on ideas. While a car will eventually run out of gas, and bacteria will run out of nutrients, ideas are resources that grow the more they are used. Existing ideas give birth to new ones, which soon begin to multiply. So, while our ability to further improve internal combustion engines, computers, or antibiotics may indeed have diminished, we now look forward to new advances in genetic engineering, regenerative medicine, nanotechnology, and artif i cial intelligence – f i elds that will once again revolutionize science and our society, opening up whole new chapters beyond our wildest imagination.
https://ivanov-p.livejournal.com/288720.html
