Tuesday, April 2, 2019

A brief outline of my road to weapons effect skepticism

I have a much longer post in the works. At the moment it is nowhere near ready for public consumption. I do think that in the meantime it might be helpful to lay down some talking points regarding my particular journey:

1. I was involved in some of the early research that showed a link between mere exposure to various weapon images and accessibility of aggressive thoughts. That's part of my history, along with my general acceptance that the Carlson et al. (1990) meta-analysis had essentially closed the case regarding the link between the mere exposure of weapons on aggressive behavior.

2. Post-graduate school, as a professor who primarily focuses on instruction, I continued to find interest in the weapons effect. Research opportunities were few and far between, of course. I continued to share what I knew about the available research.

3. At some point around the start of this decade, I got serious about updating the old Carlson et al. (1990) meta-analysis. I was reproducing the original effect sizes using a now very antiquated software called D-Stat (way too limiting - never use again). That was successful insofar as it went. I got an offer to do something considerably more sophisticated, and was promised access to better software and expertise. I could not refuse.

4. I would end up coauthoring the occasional narrative literature review that gave a glowing portrayal of the weapons effect literature. I believed what I wrote at the time as it was consistent with the findings I was involved in generating from the new meta-analysis, and generally supportive of the original Carlson et al. (1990) meta-analysis. In hindsight, I came to realize I was wrong.

5. Eventually an updated meta-analysis I coauthor gets accepted for publication. Yay. Then boo. There turned out to be a database error that I and the individual who was third author on the article never caught. The fact that we never caught it still bugs me to this day. I have emails documenting my requests to said coauthor to verify that the database was accurate over the period prior to publication.

6. Reanalyses required a rethink. Publication bias is a serious problem with this literature. Establishing a link between exposure to weapons and aggressive behavioral outcomes is difficult at best, and probably not doable. Moderators that appeared to be interesting were not so interesting.

7. How do you go about redrafting an article when each of the coauthors is working at cross-purposes? Hint: it ain't easy.

8. Bottom line: I cannot speak for my coauthors, but I cannot unsee what I saw. Based on the available evidence, I can no longer have confidence that the weapons effect is a legitimate phenomenon. That is not a knock on Berkowitz, but is rather the cold hard truth as I see it after looking at the evidence available was available.

9. Initial analyses I ran last fall after the revised manuscript was accepted show that there is also a potential allegiance effect. That really needs further exploration.

10. Although there are analyses that I certainly would love to run or wish I had run, the bottom line remains: there are likely issues with sample size and research design in these experiments that makes assessing behavioral outcomes darned difficult at best. As a social prime, short-term exposure to weapons may not be particularly interesting.

11. I honestly believed an effect was real that apparently was not. Once the evidence to the contrary socked me in the jaw (to borrow a phrase from Wittgenstein), I had to change my perspective. Doing so was not easy, but I have no regrets.

I'll lay out the longer version of this later.

No comments:

Post a Comment