Three for Thursday

Let’s dive into three refreshing topics today.

Another “Predetermined Narrative”

Earlier this week, I wrote about a bad preprint devoted to what appeared to be a predetermined “influencer” narrative — so much so that the authors ignored data staring them right in the face which suggested a more mundane but likely explanation, and clung relentlessly to the idea that individual “influencers” drive citations in the AI and ML research literature.

This was a preprint, so expectations were low going in.

Now, a peer-reviewed study in Circulation: Arrhythmia and Electrophysiology shows that pre-determined narratives can also survive peer-review. In this case, the authors contend that artificial sweeteners are linked to people developing atrial fibrillation (afib). The AHA’s press office fell all over themselves promoting the study, and the press has splashed it far and wide without question.

As a devoted Diet Coke drinker, I have had study after study about the risks of artificial sweeteners forwarded to me by well-intentioned loved ones — so I’ve learned to find the flaws. As seems the case in this instance, it usually isn’t difficult.

First, a bit of background on afib, which has a few well-known triggers — stress, alcohol, and caffeine are at the top of most lists. A lot of beverages that use artificial sweeteners have caffeine in them — Diet Coke, Coke Zero, Diet Mt. Dew, Red Bull Sugar Free, Diet Pepsi, Diet Dr. Pepper, Pepsi Zero Sugar, etc. As a result, if you want to eliminate caffeine as a potential confounding variable, you’d better account for it.

• Caffeine is a well-known issue for the heart, and not just for afib — Panera is currently being sued over its “Charged Lemonade,” a drink with the equivalent amount of caffeine in one 30 oz. serving as 10 caffeinated sodas. A 21-year-old with long QT syndrome died after drinking one.

The authors used a questionnaire to generate their data — which means they thought about what to include — but for some reason, they did not ask whether the respondents ingested caffeinated beverages. As they note in the paper:

Third, limited by the questionnaires, we could not tell whether the SSB and ASB were caffeinated, and we cannot rule out residual confounding by other unmeasured or unknown factors.

Any limits to the questionnaire were created by the researchers. So, when one of the “unknown factors” is one of the biggest known factors, it’s time to hit the pause button. Why design a questionnaire that skips such a major and realistic confounder?

Only one author on the paper is a cardiologist, and he might be a junior one — his ORCID profile shows only three papers, one of which is about Covid-19, one which is about smoking and cardiovascular health, a pretty settled question. This second paper has essentially the same authors as the current one, a list consisting of seven endocrinologists, a public health researcher, and the lone cardiologist.

Why the editors and reviewers didn’t catch the caffeine oversight is also puzzling. The authors’ narrative was so pre-set and/or their knowledge of afib so rudimentary that they skipped asking one of the most obvious questions imaginable, and then came to a rather implausible conclusion based on data with a gaping hole at the center.

I would bet that if they went back and interrogated the respondents about the amount of caffeine they ingest via these artificially sweetened drinks, they would find a strong correlation between caffeine intake and afib risk. It’s a well-known trigger, it makes sense physiologically, and it’s a clinically relevant question.

I am also going to go on record as predicting this paper gets retracted.

With caffeine in the bucket of “unknown factors,” they stuck with their narrative, turning to something that makes far less sense — artificial sweeteners. They could just have easily created a narrative that aluminum cans are associated with afib, and it would have made as much sense.

We’ll call this one “Study Zero Insight.”

Congress Asks for More Reporting from OSTP

• Correction: The first report from OSTP was issued in November, something I missed in the initial post — it rang a bell, but I couldn’t find it, even though once I saw it again thanks to an attentive reader, I remembered it in all its bloated glory. Instead of an extension as described in the initial post, this is a new deadline for an additional report. The language of this post has been modified accordingly after being sent via email. I apologize for the error.

The new stopgap spending bill has been signed, and OSTP has been asked to deliver another report in June.

Wording in the stopgap spending bill reads:

Open Access. — The agreement directs OSTP to produce an in-depth financial analysis of the August 25, 2022, Memorandum to Executive Departments and Agencies titled, “Ensuring Free, Immediate, and Equitable Access to Federally Funded Research” including the policy’s anticipated impact on Federal research investments, research integrity, and the peer review process, as was previously directed in House Report 117-395, no later than 100 days after the enactment of this act. If OSTP fails to provide the Committees with the report within 100 days, then OSTP must pause implementation of the memorandum until the agency produces the report.

This follows the initial request, which generated a report delivered in November — one with some breathtaking cost estimates.

It’s also notable that the term “public access” has been superseded in these documents by the term “open access” — a small, but potentially important shift in vocabulary.

The Real-world is More Fun

All the talk of AI and virtual this and virtual that seems to miss a major point — being human is kind of fun and enjoyable. In the world of screens, Zooms, and remote offices, it can feel like it’s all about tasks, productivity, searches, and outputs, but life is far richer than that.

The implicit argument from Silicon Valley is that there’s a rich world to be discovered online. But you have to really lower your expectations to believe it.

I’ve been looking for examples to show these contrasts, and want to share a couple recent ones. Once you start looking, you may see them everywhere in your own life, and start leaning into non-screen experiences a little more.

The Grocery Store Run

I had to go to the grocery store to pick up a couple of items. And it was more fun than ordering anything from Amazon or Peapod has ever been:

• I was able to prank a neighbor, who set down her small-items basket to bag some oranges, so I simply approached from behind and started walking off with her basket. She’s fun, so she exclaimed, “Hey!” and then laughed when I turned around, we hugged, and then spent the next five minutes laughing and catching up on things. It was nice.
• I found item alternatives I didn’t know existed, and bought these to take to an evening event instead of what I’d planned. They turned out to be mediocre, but they were different, and that was a good thing.
• I was able to help a stranger, an older guy who had a hatchback full of groceries but who had forgotten to close the hatch. He was backing out, so I flagged him down, explained why I’d done so, and then with his permission closed his hatch, saving his groceries from spilling all over the parking lot.

This all happened in about 15 minutes, and it was a nice reminder that non-screen, non-tech life is actually where you have the most fun, finds, and benefit to others.

The Mysterious Intervention

Recently, an event had the potential of becoming explosive. One friend appeared to be dating the ex-girlfriend of another friend, and was planning on bringing her to the event. This would have been news to the girlfriend’s ex, and may have led to a multi-level meltdown.

What to do? Use an AI? Do a Google search? Write a blog post?

Or delve into the mysterious world of body language, friendly persuasion, and back-channel sidebars?

While it’s not entirely clear what happened, some amount of conversation, hint-dropping, raised eyebrows, private thoughts, and so forth resulted in the new couple realizing this was not the time to reveal their new relationship. One person decided it was more than appropriate for them to stay home, and the evening went swimmingly.

This is the kind of stuff I would never trust to an AI.