[The retraction of the Lewandowsky/Cook “Recursive Fury” paper has been generally greeted as a great victory for us sceptics, and many commenters have opined that we can forget about it; it’s history. I disagree, for the following reasons.
The same defence of Lewandowsky’s scandalous pseudo-science is still being offered in the same quarters, with the additional message that he is the victim of an organised campaign of bullying. The message that climate sceptics are conspiracy theorists has entered the mainstream, and has been repeated by a British government minister and a website published under Barak Obama’s name.
Simply pointing out that something is wrong is no guarantee that it will go away. Society has never worked that way, and the fact that the way information is transmitted, challenged and retained has been revolutionised by the internet means that the transmission of information is even more of a mystery than it has ever been. Lewandowsky understands this, and so do governments, private companies (including scientific publishers) and pressure groups, who are tentatively feeling their way round, as we all are.
“Recursive Fury” is a magnificent subject for the content analysis which Lewandowsky and Cook attempted and made such a hash of. THere is material here for a book, or dozens of research papers. We have two faulty scientific papers, plus partial supplementary material. We have FOI material including correspondence between authors, blog owners, and university officials, ethical permission files, letters of complaint, letters to and from editors; we have Cook’s private assessments of Lewandowsky from the leaked “treehut” files; we have hundreds of blog articles, tens of thousands of blog comments, and scores of articles by Lewandowsky on his own blog and in the mainstream media and “scientific” literature.
In order to facilitate examination of this monumental cock-up, this “slow motion train wreck” as Anthony Watts called it, I reproduce below a summary of the paper, referenced by page numbers. There’s material here for dozens of papers; I’m working on “Lew’s Lost Conspiracy” from the supplementary material, and hope to post it in the next couple of days. I don’t expect many people will be interested in reading it all, but it’s there for anyone who wants to do further research on it. Enjoy.]
Summary of “Recursive Fury”
[All direct quotes from the paper are in italics. I’ve added some emphasis in bold to facilitate identification of sections]
“Recursive fury: Conspiracist ideation in the blogosphere in response to research on conspiracist ideation” by Stephan Lewandowsky, John Cook, Klaus Oberauer and Michael Hubble-Marriott is a long (57 pages) paper provisionally published on-line by “Frontiers in Psychology” on 2nd February 2013. It was published on 18th March 2013, revised twice following complaints, and then “removed” on 27th March 2013 pending an internal enquiry. It was finally retracted on 4th April 2014.
The paper analyses the response of the climate blogosphere to the publication of a previous paper by Lewandowsky, Oberauer and Gignac (LOG 12).
The key contents of the paper are as follows:
pp3-6 [no subheading] consists of an analysis of the concept of conspiratorial ideation and its involvement in the rejection of science, with reference to the AIDS/HIV controversy, tobacco and health, vaccination and climate science.
pp6-8 Conspiracist ideation and rejection of science among climate blog visitors: summarises the results of LOG12 thus:
“… endorsement of free-market ideology emerged as a strong predictor of the rejection of climate science. Free-market ideology was also found to predict the rejection of other scientific propositions. Of greater interest in the present context is the association between conspiracist ideation and the rejection of climate science and other scientific propositions, although the strength of this association was considerably less than that of free-market ideology.”
And goes on to say;
“When the article by Lewandowsky et al. became available for download in July-August 2012, the climate denialist blogosphere responded with considerable intensity along several prongs: Complaints were made to the first author’s university alleging academic misconduct; several freedom-of-information requests were submitted to the first author’s university for emails and documents relating to LOG12; multiple re-analyses of the LOG12 data were posted on blogs which purported to show that the effects reported by LOG12 did not exist; and a number of hypotheses were disseminated on the internet with arguably conspiracist content. […]
The remainder of this article reports a content analysis of the hypotheses generated by the blogosphere to counter LOG12. The extent and vehemence of contrarian activity provided a particularly informative testbed for an analysis of how conspiracist ideation contributes to the rejection of science among web denizens. Unlike previous analyses of web content, the present project was conducted in “real time” as the response to LOG12 unfolded, thus permitting a fine-grained temporal analysis of the emerging global conversation.”
Two searches of the internet were conducted by coauthors Cook and Marriott. The first one obtained all peer-reviewed papers on conspiracy ideation published in 2012 (up to October 12).
“The second phase of the search traced the response to LOG12 in the blogosphere. An on-going web search in real time was conducted by two of the authors (J.C. and M.H.M.) during the period August-October 2012. This daily search used Google Alerts to detect newly published material matching the search term ‘Stephan Lewandowsky’.”
The data was subsequently limited to the 30 most frequently read “skeptic” websites, as identified by Alexa rankings, and to the period 28 August to 18 October 2012.
pp10-12 Conspiracist classification criteria
lists six criteria of conspiracist ideation, with references to the literature. The authors say:
“Our criteria were exclusively psychological and hence did not hinge on the validity of the various hypotheses. […] The approach […] avoids the need to discuss or rebut the substance of any of the hypotheses.”
These are the six criteria, together with abbreviations used in the text for each one, and explanatory extracts from the paper:
NI Nefarious Intent: “..the presumed intentions behind any conspiracy are invariably nefarious.”
PV Persecution-Victimization: “… A corollary of the first criterion is the pervasive self-perception and self-presentation among conspiracy theorists as the victims of organized persecution. The theorist typically considers herself, at least tacitly, to be the brave antagonist of the nefarious intentions of the conspiracy; that is, the victim is also a potential hero.”
NS Nihilistic Skepticism: “The conspiracy theorist refuses to believe anything that does not fit into the conspiracy theory. Thus, nothing is at it seems, and all evidence points to hidden agendas or some other meaning that only the conspiracy theorist is aware of.”
MbW “Must be Wrong”: “The underlying lack of trust and exaggerated suspicion contribute to a cognitive pattern whereby specific hypotheses may be abandoned when they become unsustainable, but those corrections do not impinge on the overall abstraction that `something must be wrong’ and the `official’ account must be based on deception.”
NoA “No Accident”: “To the conspiracy theorist, nothing happens by accident … Thus, small random events are woven into a conspiracy narrative and reinterpreted as indisputable evidence for the theory.”
UCT “Unreflexive Counterfactual Thinking”: “Contrary evidence is often interpreted as evidence for a conspiracy […] the stronger the evidence against a conspiracy, the more the conspirators must want people to believe their version of events.”
After a brief discussion of their search for other papers on conspiracy ideation published in 2012, which revealed that only LOG12 engendered “recursive hypotheses”, (Table 2) the authors turn to a discussion of “at least ten […] hypotheses advanced against LOG12, irrespective of whether they addressed presumed flaws in the methodology or accused the authors of deception, incompetence, or outright conspiracies. […] We do not comment on the validity of any hypothesis other than those that can be unambiguously classified as false (namely, hypotheses 2, 6, 7, and 8).”
Each hypothesis is backed up with quotes from comments on blogs. A total of 42 links to blog comments are given, though the total number of quotes referred to may be greater.
The hypotheses are summarised in Table 3, with the appropriate conspirationist criteria in brackets:
1. Survey responses “scammed” by warmists (NI, PV, MbW, SS)
2. “Skeptic” blogs not contacted (NI NS PV)
3. Presentation of intermediate data (NI, NS, MbW, UCT)
4. “Skeptic” blogs contacted after delay (NI, NS, MbW, NoA, UCT)
5. Different versions of the survey (NI, MbW, UCT)
6. Control data suppressed (NI, NoA)
7. Duplicate responses from same IP number retained (NS, MbW)
8. Blocking access to authors’ websites (NI, PV, NoA)
9. Various Miscellaneous hypotheses (See text)
[under this heading two hypotheses were considered: 9.1 Tom Curtis’s criticisms as a false flag operation; 9.2 Moon Hoax as bait for Recursive Fury. The titles are mine, since none were given in the text.]
10. Global activism and government censorship (NI, PV, SS)
pp27-28 Freedom-of-information release
Further conspirationist hypotheses are considered under this heading. The authors say: “Because the FOI release occurred about a month after the last hypothesis spontaneously emerged in response to LOG12, it is considered separately from the other hypotheses summarized in Table 3.”
The FOI release occurred on 10th October 2012.
pp29-31 Discussion: Potential limitations
This section analysed a number of potential objections, including:
1) The “generality” of the results. The authors counter this by saying:
“We therefore suggest that the present analysis illuminated not just an isolated incident but the broader propensity of climate denial to involve a measure of conspiracist ideation.”
2) That the analysis: “…considered the ‘blogosphere‘ as if it were a single entity, analyzed within the context of psychological processes and constructs that typically characterize individuals rather than groups.”
To which they reply: “Our response is twofold: First, at the level of purely descriptive discourse analysis, our work fits within established precedent involving the examination of communications from heterogeneous entities such as the U.S. Government or the Soviet Union. Second, at a psychological level, numerous psychological constructs—such as cognitive dissonance, social dominance orientation, or authoritarianism—have been extended to apply not only to individuals but also to groups or societies…”
3) That “…the evidence falls far short of “real” conspiracy theories involving events such as 9/11 or the moon landing.”
To which they reply: “We suggest that conspiracist ideation, like most other psychological constructs (e.g., extraversion), represents a continuum that finds expression to varying extents in theories of varying scope.”
4) “..critics might propose an alternative explanation for the behavior of the blogosphere based on a dissonance effect. Science denial commonly involves “skeptics’” self-perception of being the only rational consumers of information in a sea of corrupt or self-serving scientists). […] this hypothesis is not in opposition to ours: We would expect that a person’s disposition to engage in conspiratorial thinking is more likely to become manifest when triggered by factors such as cognitive dissonance.”
5) “Critics might furthermore argue that our analysis of the response to LOG12 was over-extensive, and that some of the hypotheses advanced by the blogosphere in fact constituted legitimate criticism. This criticism is rendered less potent by the fact that our analysis was conducted at a psychological level, without regard to the truth value of any of the hypotheses other than those that could be unambiguously classified as false (i.e., hypotheses 2, 6, 7, and 8 in Table 3). We remain neutral with respect to the question whether the remaining hypotheses presented valid criticisms.”
and they add: “Our decision not to address the validity of any of the hypotheses also helps allay one important remaining issue: Two of the present authors also contributed to LOG12, and the present analysis may therefore be biased by a potential conflict of interest. This possibility cannot be ruled out […][B]ecause data collection (via internet search) was conducted by two authors who were not involved in analysis or report of LOG12, the resulting “raw” data—available in the online supplementary material—cannot reflect a conflict of interest involving the LOG12 authors. Moreover, the availability of these raw data enables other scholars to bring an alternative viewpoint to bear during any reanalyses.”
pp32-35 Theoretical and pragmatic implications: Implications for understanding conspiracist ideation.
The criteria “Must be Wrong” and “Self-Sealing” are examined in more detail, and the authors suggest: “… that some of the variables that predict conspiracist ideation—viz. low trust and paranoid ideation were observable in the response to LOG12.”
They mention: “…the well-established fact that the rejection of climate science is strongly associated with right-wing political leanings and the embrace of “fundamentalist” laissez-faire vision of the free market”and add: “One might therefore be tempted to consider conspiracist ideation another manifestation of the “paranoid style” in American politics—mainly focused on the political Right—that was famously highlighted by Hofstadter (1966).”
However: “There are several indications that acceptance of this view would be premature: LOG12 found no association between conspiracist ideation and free-market ideology […] and in a similar study involving a representative sample, Lewandowsky, Gignac, and Oberauer (2013) found conspiracist ideation to be negatively associated with free-market ideology and conservatism.”
“… we uncovered a potentially novel aspect of conspiracist reasoning when some of the later hypotheses were found to involve a residual impact of earlier, discarded hypotheses. For example, whereas critics initially argued that the results of LOG12 were invalid because “skeptic” bloggers were not contacted (hypothesis 2 in Table 3), upon release of evidence to the contrary, the same conclusion of invalidity was reached by other means; either because of a preliminary report of the data during a colloquium (hypothesis 3); or because of the presumedly faulty timing of the correspondence (hypothesis 4); or because “skeptic” bloggers were emailed different versions of the survey (hypothesis 5). All of those hypotheses rely on counterfactual thinking because no “skeptic” blogger posted links to the survey, and therefore neither the dates of correspondence nor the version of the survey (nor any other event involving those bloggers) could have affected the data as reported in LOG12.”
The third and final implication for understanding conspiracist ideation is that:
“There are other streams of science denial that are detectable in the response to LOG12. For example, the repeated re-analysis of data, involving the elimination of “inconvenient” subsets of data points based on fairly fluid criteria...”and a parallel is drawn with reanalysis of health statistics by scientists working for the tobacco industry.
pp35-36 Theoretical and pragmatic implications: Implications for understanding science denial.
The authors say:
“The vast majority of domain experts agree that the climate is changing and that human CO2 emissions are causing this change. Given this broad agreement on the fundamentals of climate science, what cognitive mechanism would underlie people’s dissent from the consensus? […] Rejection of the scientific consensus thus calls for an alternative explanation of the very existence of that consensus. The ideation of a secretive conspiracy among researchers can serve as such an explanation. Moreover, the ideation of a conspiracy may also serve as a ‘fantasy theme‘ that permits groups to develop and share a symbolic reality. […] Fantasy themes are known to play a major role in climate denial.”
pp36-37 Theoretical and pragmatic implications: Implications for science communication.
“Although suggestions exist about how to rebut conspiracist ideations […] we argue against direct engagement for two principal reasons. First, much of science denial takes place in an epistemically closed system that is immune to falsifying evidence and counterarguments. We therefore consider it highly unlikely that outreach efforts to those groups could be met with success. Second, and more important, […] [a]lthough LOG12 received considerable media coverage when it first appeared, the response by the blogosphere was ignored by the mainstream media. This confinement of recursive hypotheses to a small “echo chamber” reflects the wider phenomenon of radical climate denial, whose ability to generate the appearance of a widely held opinion on the internet is disproportionate to the smaller number of people who actually hold those views […] Thus, although an understanding of science denial is essential given the importance of climate change and the demonstrable role of the blogosphere in delaying mitigative action, it is arguably best met by underscoring the breadth of consensus among scientists rather than by direct engagement.”
p48 Author Note
Table 1 simply lists the ten most popular climate sceptic blogs, six of which are quoted in the paper.
Table 2 lists the twenty papers on conspiracy theories published in 2012, together withy the total number of Google hits (443 for LOG12, and 124 for all the other 19 papers) and the number of recursive theories (10 for LOG12, 0 for the other 19 papers).
Table 3 lists the ten conspiracy theories, and attributes to each one the relevant conspiracy ideation criteria, as indicated in the discussion of Results pp13-37 above. It also gives the date on which each hypothesis was first proposed, and the name of the person proposing it. Four hypotheses are attributed to Steve McIntyre, two to Joanne Nova, one each to Anthony Watts, Geoff Chambers and “ROM”, and one to “Various”.
Figure 1 is the latent variable model from LOG12 presenting the results of that paper in diagrammatic form.
Figure 2 is the timeline of principal recursive theories developed by the blogosphere in response to LOG12. All comments were collected between 29th August and 23rd September, except for two comments collected on the 14th and 18th of October.
Numbers of comments per day and per hypothesis are recorded in different shades of grey, which makes counting difficult, but there see to be total of 62 comments recorded.
The Supplementary Material was placed on-line soon after the paper. It consists of a spreadsheet, with dates along the side and conspiracy theories along the top, with quotes from blog comments in the corresponding cells.
The conspiracy theories listed were:
Didn’t email deniers (29)
Inconsistent delivery/excluded skeptics (4)
Warmists faked data (38)
Methodology flaws (35)
Emailed warmists before deniers (4)
Intermediate Data (8)
SkS conspiracies (15)
Hiding Data (6)
STW Censoring Comments (5)
Used multiple IPs (4)
Kevin Judd puppet master (3)
Tom Curtis faked criticism (1)
SL founded Conversation (1)
Lew gravy train (1)
Blocks IPs (2)
97% of deniers didn’t enter survey (1)
Paper isn’t going to be published (1)
Unethical ethics application (3)
Gravy train (1)
Govt conspiracy theory (1)
(numbers of quotes for each conspiracy are listed in brackets)