Jens’ Bayesian Models

[W] Hello. How’s everything?

[J] Things are good, thanks 🙂 Busy December, but that’s par for the course. I hope you’re well also.

[W] I have the flu, but hope to survive against all odds. That’s you:

[J] Ah, the classic December curse. Yep.

[W] And here’s a video of you

[J] Yeah, it was a book launch a month and a bit ago

[W] Wow, you can talk! Wait – you launched a book?

[J] Yeah, the talk was a presentation of the book.

[W] Ahhhhhhhhhhhhhhhhh. OK. Now, let me get this straight. You published a book in 2019, and what, 88 articles?

[J] Yes – the book was published in 2019. Haha, hardly – I think I published 7 papers in 2019 or so

[W] So, tell me – how many hours does writing an article take you?

[J] How many hours? That depends on the length and technical difficulty of the paper – usually, it probably takes a few days or so to draft it, but with model development, testing, data collection, analyses, revisions, and contributions from co-authors, it can easily take a few months from the initial idea to the submitted paper. But once the legwork is done, usually the draft takes a few days.

[W] Give me a ball park in terms of expectation. I tell you – Jens let’s write a thing, how many hours you need to put aside. Two full weeks?

[J] It entirely depends on this [what you’re going to write] – for example, smaller papers can take 1-2 days while larger papers can take weeks 🙂

[W] So about a full week. You’re good. First article you wrote is this.

[J] Yes, that’s the earliest paper published in 2019, but I’ve been publishing papers since 2012 🙂 I think my first paper was this: Because Hitler did it! Quantitative tests of Bayesian argumentation using ad hominem

[W] Oh, I love ad homs. I know you have tricks, I just want readers to understand they’re obsessing over something that happens in a limited time span.

[J] That makes sense – managing reader expectations is crucial.

[W] I’m working on the practical aspects of fallacy fluff these days. I think people misunderstand them, and that it’s the philosophers’ fault.

[J] Oh cool – there’s a lot of interesting Bayesian papers on fallacies.

[W] You work on Bayesian models, that seem the constant in your research.

[J] Yep – I’ve been working a lot with Bayesian models 🙂

[W] OK, which paper should we look into, your most recent one?

[J] I figured the seepage paper was the one that sparked your interest, yes?

[W] Yes. There is also this one but I must admit that it’s more you that interests me. I want to show a variety of scientists. Here’s the seepage paper:

We present an agent-based model of scientific consensus formation in climate change.

An evidence-resistant minority can prevent the public from acquiring the consensus position.

An evidence-resistant minority can delay, but not prevent, a scientific consensus.

The model matches several aspects of public opinion formation and consensus formation

Love the hightlights.

[J] Thanks 🙂 We felt the model in the paper yielded some interesting perspectives on belief revision and the impact of contrarian interpretation of data

[W] Let’s start with the basics, what’s a Bayesian model?

[J] It’s a probabilistic model that integrates two things: prior beliefs and some new evidence (which can be more or less relevant to the hypothesis under consideration). From these, Bayes’ theorem gives you the posterior degree of belief in the hypothesis given that evidence. Critically, Bayes’ theorem can use subjective probabilities – that is, two people may have different priors and thus arrive at different posterior degrees even when given identical information

[W] Good point. So it’s how people should in principle revise their beliefs based on new evidence. Your model says that contrarians can’t prevent correct beliefs to spread.

[J] Our simulations showed that unbiased agents necessarily acquire belief in the climate-change hypothesis, even when they start from an initial position of extreme skepticism and even when they rely on un- duly short temperature trends.

[W] Wow. But then we’re all biased, no? Even I am biased toward your research.

[J] We use ‘agents’, as we use an agent-based model (‘agents’ are just the individual actors in the simulation). To some degree, we probably all are biased to some extent by our personal experiences.

[W] So contrarians can’t prevent correct beliefs to spread, but you found that some contrarians are unconvinceable, right?

[J] The contrarians in the model are people who transmit biased interpretations of the data to the population. But yes, we show that even the contrarians cannot remain sceptical without biased interpretations of the data.

[W] But you can’t predict that contrarians won’t cling to their priors. You’re not saying that correct information can win against the staunchest of contrarians.

[J] True – but in the model, we wanted to see what happened when they updated their beliefs. An alternate model could simply have contrarians who only transmit their initial prior without developing or changing their beliefs over time.

[W] That would go against my experience. Contrarians usually move into a matrix. They can also play at what I call the climateball bingo.

[J] The idea of not moving their belief could be integrated within the model – in this paper, however, that wasn’t what we wanted to explore 🙂

[W] For instance: But1940, But2ndLaw, But70s, ButABC, ButAbsoluteTemps, ButAcid, ButActivism, ButAdjustments, ButAdvocacy, ButAl, ButAlarmism, ButAnonymous, ButAntarctica, ButArrhenius. And that’s just for before B. Contrarians move around the talking points.

[J] Do you refer to data cherrypicking? Ah, I see – moving the goal post

[W] There are many ways, yes. You can doubt anything but data, or some dataset, or prefer that other data set, or etc. It never ends. It’s quite subtle, contrarians are quite good actually at ClimateBall bingo.

[J] For cherrypicking, The analysis found that since 1970, any bet against warming—even those involving cherry-picking of short-term cooling trends—would have been unsuccessful in the model that is 🙂

[W] Ah, that’s a good point.

[J] Yes, cherry-picking and moving the goal post is really pernicious. One of the points of the paper is to show that cherry-picking cannot yield denial, but that the contrarian must have some biased interpretation on top of cherry-picking. The model used on a two-pronged propagandistic effort: first, promoting and sharing of independent research that conformed to the industry’s position, and second, funding of additional research with selective publication of the results. This was when we wanted to test the influence of contrarians on public opinion

[W] How would you characterize that influence, e.g. is it worth their time to create FUD, i.e. Fear, Uncertainty, Doubt?

[J] The emotional state of the recipient probably influences their perception of the argument and data, but unfortunately I’m not well-placed to say how, as my research hasn’t been concerned with emotions. However, there’s some excellent research done on emotions and reasoning by people like Isabelle Blanchette and others

[W] Will look. Even from an epistemic standpoint, you’re saying that promoting tendentious research slows down the correct information to spread, but in the end it’s insignificant, right? I’m wondering why one would waste time promoting tendentious research.

[J] It might not be insignificant if contrarians are given a large platform.

[W] And they are.

[J] Also, the contrarians might have different incentives than epistemic accuracy – e.g. lobbyists who argue for a very particular point.

[W] The industries are controlling most media. Just look at how Brexit got covered:

[J] Most media are industries in and of themselves 🙂 Yeah, it’s pretty messy – a large part of my research is concerned with how information spreads on social networks, e.g. how echo chambers emerge.

[W] It’s hard to deny that people got served Corbyn’s antisemitism. Not saying it’s not worth discussing, but I think there are other interesting things we could have discussed.

[J] Yeah, that message seems to have been pretty consistent.

[W] I need to ask you about Brexit. You’re Danish. Will it affect you?

[J] I’ve got settled status, so it probably won’t affect me personally (although, I am of course impacted by the economy if that turns bad). But it’ll leave a lot of people in uncertain positions

[W] Good to hear. People around you?

[J] Yeah, friends and colleagues who don’t know if they can (or want) to continue to live in the UK.

[W] Damn.

[J] I guess that’s to be expected, though.

[W] OK. You have a gig, a book, and papers, and… music? Tell me you like music.

[J] Of course 🙂 who doesn’t like music?

[W] Exactly why I ask. What should I listen to absolutely in the next 48 hours

[J] How about some Tom Waits?

[W] Great

[J] Depends on your mood and tastes, I guess.

[W] Which album?

[J] Real Gone? At least, I like that album 🙂

[W] Good choice, I like Alice. And which Danish artist do you think the world would need to know more.

[J] Malk de Koijn (Danish rap group).

[W] Video?

[J] Their lyrics are hilarious (although, unfortunately in Danish).

[W] Tack så mycket, Jens

[J] You’re welcome 🙂

About Willard
This entry was posted in Uncategorized. Bookmark the permalink.

9 Responses to Jens’ Bayesian Models

  1. Willard says:

    The Seepage paper is here:

    (I would never suggest to go find it on a scientific hub somewhere. Never.)

    Interestingly, Tom Waits reached the 70th milestone this month. A recent poem from him dedicated to Keith:

    We’re all getting older, except for Keith Richards.

  2. Robin Williams on Keith Richards:
    “I know there is a cure for whatever bioterrorism they throw at us, I know there’s one. And it lies within Keith Richards, I do know that. He is the only man on the planet that could go (imitates Keith snorting something) “Anthrax? All right! Doesn’t go with my E. coli, but fuck.” Keith is the only man who makes The Osbournes look fucking Amish. He’s insane. I’ve seen Keith go to a drug dealer and the drug dealer’s goin’, “I’m out, man, I’m sorry. I have nothing left!” Supposedly, he goes to Switzerland and changes his blood. Not like one pint, but like a fucking Chevrolet, all of it. I just wanna know, who gets his blood? Some old Swiss man’s going, “HEIDI!! We’ve gotta go on tour, you bitch! We’ve gotta pay for Mick’s babies! COME ON!!!” Because I know this. I know that we may all be dead and gone, Keith will still be there with 5 cockroaches. Keith will go, “Ya know, I smoked your uncle! Did ya know that? Fucking crazy!”

    I enjoy your posts, they make me think in ways different from the norm. Perspective changes with the accumulation of knowledge from thought provoking source material.

  3. Willard says:

    Thank you for the kind words, A.M.!

    Merry Festivus!

  4. You are very welcome. 🙂

    Merry Festivus to you as well!

  5. David B. Benson says:

    I suppose everyone here knows that with enough evidence the posterior pdf, probability density function, becomes essentially independent of the prior pdf, the so-called subjective pdf, over the domain of support, i.e., wherever not zero.

  6. Steven Mosher says:

    [J] How about some Tom Waits?

    best . ever. Moshpit approved.

    thanks for reminding me how much I like Tom Waits.

    It was 1981, I was spending a lot of my time off campus with my frat brother.
    drinking smoking and then one night he put on Waits.
    We did not discuss bayes. he was a threatre type.

    he hasnt changed much.. less hair but still the same old knepper

  7. Steven Mosher says:

    what a beautiful find Willard. Rhetorical theory, Bayes, modelling. and the dude like Waits.
    watching his presentation..

  8. Everett F Sargent says:


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.