如何正確地質疑科學

「全球變暖是偽科學，相關數據都是被篡改了的！」

「Global warming is based on faulty science and manipulated data.」

皮尤研究中心的一份報告顯示，越來越多的美國人都在過去五年都在附和著這種對科學的不信任。

A Pew Research report revealed that the American public has increasinglyechoedthis sort of scientific mistrust over the last 5 years.

誠然，科學並不是完美的。但理性地質疑科學，與因為科學發現與個人觀點相衝突就完全拒絕接受科學之間，還是有區別的。人們通常拒絕科學結論，僅因為這些結論和他們的個人世界觀相矛盾，於是他們固化了自己對科學的不信任（就算這種不信任是沒有根據的），從而他們也會傳播錯誤的科學信息。

Admittedly, the scientific process isn』t perfect. However, there is a difference between a reasonable mistrust of the scientific process and an outright rejection of all scientific findings that clash with personal beliefs. Too often, peoplespurnscientific conclusions because they contradict personal worldviews, perpetuating anunfoundedmistrust in science and the spread of scientific misinformation.

The psychology of disbelief

懷疑的心理學機制

拒絕相信科學證據在很大程度上是一種心理學現象，而且這種拒絕往往是有選擇性的，而非全局性的。想像一下如下場景：質疑疫苗接種的人通常都會引用Andrew Wakefield做的那項臭名昭著的研究，為了支持自己的觀點，他將自閉症和和疫苗聯繫在了一起。質疑全球變暖的人還會引用一篇Roy Spencer和John Christy在1990年完成的論文，他們認為證實全球變暖的證據不足。這聽上去很矛盾，就是如果這幫「反科學分子」會引用科學期刊的話，這說明他們其實是相信科學證據的。當然更有可能的是，他們只會在科學證據和他們自己心中的世界觀相吻合的時候，才會選擇相信。

The rejection of scientific evidence is largely a psychological phenomenon and is often selective rather than universal. Consider these scenarios: vaccination critics quote Andrew Wakefield』sinfamousstudy linkingautismto vaccines to support their stand, and climate change skeptics cite a 1990 research paper by climatologists Roy Spencer and John Christy arguing for the lack of evidence for global warming.Paradoxically, quoting scientific journals demonstrates that these 「anti-science」 folks do actually trust scientific evidence, but probably only if italigns witha mental model of the world they have built.

對於這個現象，現代神經科學已經給出了一些解釋。當人們面對和自身想法相悖的結論時，背外側前額葉，即負責壓迫自身不想要的想法的大腦區域，會變得異常活躍。這個區域也會在你接受史楚普實驗（圖1）時變得很活躍。這種對不協調信息的壓迫，也許可以解釋為什麼人們會拒絕接受和自己根深蒂固的想法相矛盾的觀點，哪怕新觀點是正確的。畢竟，接受和自己的世界觀一致的觀點是更容易的，因為這無需耗費額外的（且是讓人不快的）腦力去調和矛盾。

As it turns out, modern neuroscience hasshed some light onthis phenomenon. When people are confronted with an idea that contradicts their beliefs, a brain area responsible for suppressing unwanted representations, the dorsolateral prefrontal cortex, becomes activated. This area is also activated when you takethe Stroop test(Figure 1). This suppression of discordant information may explain why people reject ideas (even correct ones!) that contradict entrenched beliefs, because it is easier to accept something that aligns with their own worldview than to expend the (unpleasant) extra mental effort toreconcilethe conflicting ideas.

然而，僅僅因為新觀點和自己的想法不一致就懷疑科學證據，這並不會減少科學的可信度。我們真正要問的問題是：我們怎麼才能知道那些「冒犯」我們的科學證據就一定是對的呢？

However, mistrusting scientific evidence on the basis of beliefincongruitydoesn』t make the science any less true. The real question we should be asking is: how do we know if the 「offending」 piece of evidence is actually true?

圖1：史楚普實驗: 圖中上面那欄顯示了與其字體顏色一致的單詞，而下面那欄顯示了與其字體顏色不一致的單詞。現在你要試著說出兩欄中每個單詞分別是什麼顏色，並給自己計時。不出意外，你在說下面那欄單詞的顏色時會花更長時間，因為你需要調和單詞的含義和顏色。

Figure 1: The Stroop test: the upper panel shows words that have concordant colors, while the bottom panel shows words that have discordant colors. Try to name the color of the words in both panels while timing yourself. Chances are, you will take longer to complete the second panel because you have to reconcile the conflict between the colors and words.

Mistrusting science: a scientist』s perspective

科學家是如何質疑科學的

科學給公眾展示的通常都是最終成果；之前的實驗流程和論文發表流程只有神秘的科學圈自己才知道。然而科學家在「黑箱」里到底操作了什麼，對於取得公眾看到的最終成果是非常重要的。

Scientific evidence presented to the public is often only the end result; the experimentation and publishing process is a mysterious business privy only to the esoteric scientific community. Yet, what occurs within this black box is crucial to shaping the results that the public receive.

所以，「黑箱」裡面到底發生了什麼？

So what happens in the 「black box」?

當科學家有了足夠的實驗數據，他們會把稿件交到一份學術期刊（圖2）。然後期刊的編輯部會評估這份稿件，並把稿件發給評議人（即「同行評議」階段）。作者會收到評議人的意見，並根據意見修改自己的稿件，最後再重新提交給期刊發表。

When scientists have enough experimental data, they will submit a manuscript to an academic journal (Figure 2). This manuscript is then assessed by the editorial board and sent out for review (the 「peer-review」 process). The authors receive reviewers』 comments, refine their manuscript, and resubmit it for publication.

圖2: 實驗和論文發表流程：實驗數據被整合到稿件中，再背提交到期刊。編輯會把稿件發給從事相關領域研究的評議人進行評估，評估的方面包括數據、質量和成果的影響。反饋內容將被回復給原作者，原作者會根據評議人的意見修改並重新提交稿件。如果期刊接受了作者的修訂，那麼這份稿件就可以發表了。

Figure 2: The experimentation and publishing process: experimental data is compiled into a manuscript, which is submitted to the journal. The editors send the manuscript to reviewers working in related fields for assessment of data quality and impact of findings. The feedback isrevertedto the author, who addresses the reviewers』 comments in the next submission of the manuscript. If the revisions are accepted, the manuscript is then published.

Where experimental analyses can go wrong

實驗分析也有出錯的時候

能被發表的論文，通常都能在其所在領域中掀起軒然大波。所以為了嚴謹地判斷某成果是否真的有重大突破，我們一般會使用一個叫做「P值」的工具。簡單來講，P值是用來衡量某個「零假設」為真的可能性的。所謂零假設，就是極端保守地假設兩個變數或條件並無實質差異（比如假設用藥物治療和不用任何藥物治療並沒有什麼區別，儘管我們是想驗證藥物的療效）。當P值超過了某個預先設定的閾值時，比如兩個變數之間的差異太大了以至於無法用巧合來解釋，那麼我們就能拒絕接受之前設定的零假設。通過拒絕接受零假設，我們就是在自信地說，討論中的兩個變數並不是沒有差異的。因此，我們就能下結論說，我們所觀察到的某個療法的效果是真實存在的，它不是隨機事件。我們管這叫「統計學意義」，或叫積極數據。

Often, papers that get published are the ones predicted to be game-changers in their field. To determine whether findings are ground-breaking, we commonly use something called the p-value. Simply put, a p-value is a measure of how likely it is that your 「null hypothesis」 is true. The null hypothesis is the premise that there is no true difference between two variables or conditions (i.e. treatment with a new medication vs. no treatment). When the p-value exceeds a certain pre-setthreshold, which can occur when the difference between two variables is likely too large to be explained by chance, then we can reject the null hypothesis. By rejecting the null hypothesis, we are saying, with a certain level of confidence (as dictated by the p-value), that there isn』t no difference between the variablesin question. Thus we can conclude that the effect we observe with the treatment is real and it』s not due to random chance. We call this 「statistical significance」, or positive data.

但這裡有個含混不清的地方：統計學意義並不總暗示著真實世界的意義。舉例來講，你可能會發現某個基因和精神分裂症有顯著關聯，但是如果這個基因僅僅讓發病率提高了0.2%，那就很難說這個數據是具有生物學意義的。相反，由於在大數據中的數據量太大了，具有統計學意義的關係有可能碰巧在兩件毫不相干的因素間產生——比如有項研究*發現，巧克力食用量高的國家，獲諾貝爾獎的人數也較多！

Here』s where it gets fuzzy: statistical significance doesn』t always imply real-world significance. For instance, you may find a gene that is significantly associated withschizophrenia, yet if this gene only contributes 0.2% to the disease, it can hardly be deemedbiologically significant. Alternatively, because of the sheer number of data points in large datasets, statistically significant relationships might arise between 2 completely unrelated factors by chance – like the rate of chocolate consumption and the number of Nobel laureates in a country, as this study* found!

*https://blogs.scientificamerican.com/the-curious-wavefunction/chocolate-consumption-and-nobel-prizes-a-bizarre-juxtaposition-if-there-ever-was-one/

儘管有這些問題，統計學意義還是打開期刊大門的金鑰匙。由於期刊對那些沒有統計學意義的數據持有偏見，這就刺激了篡改數據和「摘櫻桃」的行為，即有人會選擇性地呈現可以驗證假說的積極數據，並把消極數據都隱藏起來。這些行為會降低科研的效率，因為很多實驗室都會把時間和資源浪費在重複別人已經做過的實驗身上。

Despite these issues, statistical significance is often the golden ticket into journals. The bias against non-statistically significant (『negative』) data sometimes incentivizes data manipulation and 『cherry-picking』, where one selectively presents positive results that support a hypothesis, while hiding those that don』t. These practices tend to make science inefficient, as multiple labs may waste time and resources unknowingly repeating experiments that other labs have already tried.

幸運的是，越來越多的科學界人士已經意識到了這個問題。最近，世界衛生組織開始呼籲要發表所有的臨床試驗——即使是數據消極的試驗——以對那些那些失敗的試驗也做出肯定，因為它們對探索治療方法也是有影響的。今天，甚至有少數期刊*專門發表負面數據。這些新的動向將很有希望解決過分重視P值的問題，但要讓負面數據完全不再蒙羞，可能還需要一段時間。

Thankfully, there is a growing awareness of this problem within the scientific community. The World Health Organization recently called for the publication of all clinical trials – even negative results – in recognition of how failed trials can stillinformtreatment. Nowadays, there are even some journals* that publish only negative data. These new movements are promising in addressing the problems arising from an overemphasis on p-values, but it may take some time to abolish the stigma surrounding negative data.

*https://www.elsevier.com/connect/scientists-we-want-your-negative-results-too

The publishing process

發表流程

在科學領域，我們經常開玩笑說「要麼成功發表，要麼就此沉淪」。但其實在現實中，這真的是一個讓人不爽的事實。期刊可以影響公眾對科學的了解，因為它們可以在數據量巨大的稿件中，選擇發表哪些。高質量的期刊不少，但是最有名的那幾個——比如《自然》、《細胞》和《科學》——都傾向於讓更廣泛的讀者接受自己的內容。有時候，這些期刊可能會發表一些並未完全被闡明的成果——或是有爭議的成果——甚至會促使針對某些成果的進一步研究。長此以往，這種方式將讓某領域的研究更加聚焦，但是也因為這個原因，我們評估新成果的難度將更大，因為我們還需要考慮這個成果是因為什麼原因被發表的。

In science, we often joke about having to 「publish or perish」. But in reality, it is an uncomfortable truth. The sheer number of submissions that journals receive allows them to influence the type of science that reaches the public, because they get to choose what to publish. There are many high-quality journals, but the more prestigious ones – like Nature, Cell and Science, for example – tend to reach a wider audience. Sometimes, these journals may publish findings that have not been fullyelucidated– or are controversial, even – tospurfurther research on that topic. In the long run, this approach will produce a morecohesivepicture of the topic, but it complicates how you evaluate a finding because now you』ll also have to consider why a study is published.

為控制稿件的質量，我們採用了同行評議的方法——所有稿件都由該領域的專家評估。一般來講，同行評議是「單盲」的：評議人是知道作者是誰的，但反之並不亦然。這會減少一些偏見，因為評議人在批評自己朋友、競爭對手或是權勢人物的作品時，不用擔心這對自己的人際關係或職業發展有任何的影響。但是，這並不是一個萬無一失的方法。根據設計的初衷，評議人是基於他們在類似領域中的研究而被選出來的，因為這樣就意味著他們對這個領域足夠了解，以更好地判斷新發現的精度及影響。但這也可能會導致一些問題，因為有時評議人的主觀看法會影響了他對手稿中數據質量的判斷。也就是說，那些支持作者觀點的評議人更有可能建議發表稿件，但如果評議人支持與手稿理論相競爭的理論，那相反的情況就可能發生了。由於這個原因，有的期刊允許作者將他們的競爭對手排除在評議人之外，這對作者將會更公平，但也潛在地影響了評議的嚴謹程度。

When it comes to quality control, we have the peer review process – where all manuscripts are evaluated by field experts (Figure 2). Typically, peer reviews are single-blind: the reviewers know who the authors are, butnot vice versa. This removes bias by allowing reviewers to critique works of friends, competitors or powerful figures without fear of personal or professionalrepercussions. However, this process isn』tfoolproof. By design, reviewers are chosen based on their work on a related topic, because this likely means that they know the field well enough to judge the accuracy and impact of the new findings. This can lead to problems, though, if a reviewer』s own ideas cloud his judgment of a manuscript』s data quality. Those in support of the authors』 ideas would likely recommend publication, butthe reversecan occur if a reviewer supports a competing idea. For this reason, some journals allow authors to request exclusion of competing reviewers, which removes some unfairness to the author but potentially at some expense ofrigor.

總的來講，科研流程有很多微妙的細節，也很複雜。雖然其中包含了充足的保險裝置以確保發表手稿的質量，但因為系統本身的瑕疵，那些不太重大的成果有時候可能會從這些縫隙中溜走。

In sum, the scientific process is nuanced and complex, and although it generally includes sufficientfail-safesto ensure quality publication, systemic flaws do occasionally allow less robust findings to slip through the cracks.

所以，既然科學有這麼多問題，我們為什麼還要相信它呢？

So why can we still trust science despite its systemic flaws?

Trusting science: a scientist』s perspective

從科學家的角度看為什麼要相信科學

牛頓曾說：「如果我看得更遠，那僅僅是因為我站在巨人的肩膀上。」這句話正凸顯了科學的自我批判和自我修正的性質，因為新發現都依賴於之前的研究是否正確。科學家經常重複彼此的實驗，所以一個錯誤的發現想要長時間經受住這樣的審查，是非常困難的。科學中的欺詐行為——比如數據篡改或同行評議造假——是非常可恥的，不過讓我們感到欣慰的是，這種行為還是可以被其他科學家發現：這就是說，即使一項欺詐性的研究通過了同行評議，嚴謹的科學共同體遲早也會發現數據中的問題。

Newton said 「If I have seen further, it is only by standing on the shoulders of giants.」 This statementunderscoresthe self-critical and self-correcting nature of science, because further discoveries are dependent on the truth of previous research. Scientists often repeat one another』s experiments; hence it is difficult for an erroneous finding to stand up to this kind ofscrutinyfor long. Cases of scientific fraud – like data fabrication or cheating on peer reviews – are disgraceful, but the fact that they are discovered by other scientists is reassuring: it means that even if a fraudulent study makes it through peer review, the larger science community isrigorousenough to discern fallacious data.

在持續發表的大量新發現面前，我們很容易感到困惑。因此，我們有必要知道人們為什麼現在如此地不信任科學——到底是因為個人的認知偏見呢，還是因為今天高壓的政治環境呢——因為在理解科學弱點的同時，如果我們還能意識到背後的原因，那麼這將幫助我們在正確的時間，基於合理的理由質疑科學。

It』s easy to be confused by the constant slew of new findings, some of which may contradict previous studies. It is therefore more important than ever to recognize where one』s mistrust of science is coming from – especially if it stems from a personal cognitive bias or from today』s politically super-charged environment – because this awareness, together with an understanding of weaknesses in the scientific process, will help one wisely discern when and why they mistrust science.

*本文原作者Sam Wong是哈佛大學醫學院的博士在讀生。她的研究方向是癌症中的脂肪代謝，但她對科普也很感興趣，尤其喜歡從心理學角度研究和科學相關的社會議題。

*Sam Wong is a PhD student in the Biological and Biomedical Sciences Programme at Harvard Medical School. She studies fat metabolism in cancer, but is also interested in science education and the psychological aspects of social issues in science.

- THE END -

喜歡這篇文章嗎？立刻分享出去讓更多人知道吧！