Comment Permalink

JustinMElms9y10

Thanks for weighing in, Elo. I have learned from this that sometimes providing a concrete example for an abstract problem can be so distracting as to almost completely obscure the problem.

Have you tried bayesian updates?

Yes, this has been the crux of my difficulty. I have done my best to follow Bayes Theorem, my prior probability is not a strong factor (I would not be exceptionally shocked one way or the other on this particular issue, so I put my prior probability at 60% for one side), and when I get to evidence updates, I basically only have two decent pieces of evidence "Scientific organizations X, Y, and Z (of C credibility) hold this position" and "Scientific organization A and B (of D credibility) hold this other position." And then I have "The argumentation for this position is more flawed than this other position."

That seems to be just about as far as I can get, insulated from direct observation or--as you recommend--experimentation. So I am able to calculate my posterior probability and have some confidence in my approach, but I can't help but feel unsatisfied about the scope of evidence that brought me to change my position.

Also:

GMO has been around for ~10-20 years now. We are yet to see negatives of the scale predicted by the opposition to GMO. Waiting long enough has yielded evidence of absence of risk.

That is strong evidence that GMO does not have observable risks within 10-20 years of adoption, but it is considerably weaker evidence about what GMO adoption looks like after 30, 40, 50 years or a lifetime.

Lumifer9y20

That is strong evidence that GMO does not have observable risks within 10-20 years of adoption, but it is considerably weaker evidence about what GMO adoption looks like after 30, 40, 50 years or a lifetime.

You can replace "GMO" in this sentence with a lot of things. For example, "kiwi". Or "cell phone".

See in context

4 Rationality when Insulated from Evidence

by JustinMElms

29th Jun 2016

2 min read

4

Basically: How does one pursue the truth when direct engagement with evidence is infeasible?

I came to this question while discussing GMO labeling. In this case I am obviously not in a position to experiment for myself, but furthermore: I do not have the time to build up the bank of background understanding to engage vigorously with the study results themselves. I can look at them with a decent secondary education's understanding of experimental method, genetics, and biology, but that is the extent of it.

In this situation I usually find myself reduced to weighing the proclamations of authorities:

I review aggregations of authority from one side and then the other--because finding a truly unbiased source for contentious issues is always a challenge, and usually says more about the biases of whoever is anointing the source "unbiased."
Once I have reviewed the authorities, I do at least some due diligence on each authority so that I can modulate my confidence if a particular authority is often considered partisan on an issue. This too can present a bias spiral checking for bias in the source pillorying the authority as partisan ad infinitum.
Once I have some known degree of confidence in the authorities of both sides, I can form some level of confidence in a statement like: "I am ~x% confident that the scientific consensus is on Y's side" or "I am ~Z% confident that there is not scientific consensus on Y"

Once that establishes a baseline on an issue, I am able to do some argumentation analysis to see what arguments each side has that simply should not be included in the discussion. This is usually irrelevant appeals (e.g.: In the GMO labeling debate, "It must be better because it's more natural") or corollary citations that are screened off by evidence closer to the source (e.g.: In the GMO labeling debate, "X many countries require GMO labeling" should be screened off by looking at the evidence that led to that decision).

After that, I find myself with a rather unfulfilling meta-assessment of an issue. I fear that I am asking for a non-existent shortcut around the hard solution of: "If an answer is important to you, do the necessary learning to at least be able to engage directly with the evidence," but I will ask anyway: does anyone else have strategies for seeking the truth while insulated from direct evidence?

Personal Blog

4

New Comment

60 comments, sorted by

top scoring

Click to highlight new comments since: Today at 5:08 AM

[-]Elo9y90

Quick option: Look for other forms of agents you trust. Friends you can outsource some of the research to. Public figures who you can have confidence in their ideas.

I hit the same problem when trying to work out climate change circa 15 years ago. Both sides were quoting different results from the same data sources. I had no idea who to trust.

In the end of that particular problem I decided that it didn't matter to the extent that it would affect my life either way.

In this case - Can you trust the local regulatory authorities to maintain your interest enough to keep you safe insofar as you can eat anything that hits the shelves but maybe not that new experimental kickstarter-soylent-product (for now).

Also worth pointing out for this instance only: GMO has been around for ~10-20 years now. We are yet to see negatives of the scale predicted by the opposition to GMO. Waiting long enough has yielded evidence of absence of risk.

Other options that come to mind: Have you tried bayesian updates? Write down your original position, stack information on each side and see how strongly you update each way. Then decide if that is enough for you.

(3) Can you devise a test or experiment for the beliefs that you want to hold. Obviously you don't have the rigour or the time to prove or disprove the risks associated with GMO on your own. But maybe there is something you can do on your own. Even finding a test-able premise, then reading papers relating to it. Is better than nothing.

[-]JustinMElms9y10

Thanks for weighing in, Elo. I have learned from this that sometimes providing a concrete example for an abstract problem can be so distracting as to almost completely obscure the problem.

Have you tried bayesian updates?

Also:

GMO has been around for ~10-20 years now. We are yet to see negatives of the scale predicted by the opposition to GMO. Waiting long enough has yielded evidence of absence of risk.

[-]Lumifer9y20

That is strong evidence that GMO does not have observable risks within 10-20 years of adoption, but it is considerably weaker evidence about what GMO adoption looks like after 30, 40, 50 years or a lifetime.

You can replace "GMO" in this sentence with a lot of things. For example, "kiwi". Or "cell phone".

[-]JustinMElms9y00

Yes, I would agree. And I completely assent that 20 years of evincible safety can be extrapolated into "long term" (however you define that) safety more than 10 years could be. My only position in saying the above is to highlight that "It's seemed safe so far" doesn't necessarily prove that to be safe in perpetuity.

Surely you are not arguing that 20 years is the magic asymptote at which safety rises to infinity?

[-]Lumifer9y00

safe in perpetuity

Ain't no such animal.

I don't think that looking for forever guarantees is a useful exercise. The point was really the double standard applied to GMOs.

[-]JustinMElms9y00

I agree: there is no "forever guarantee," especially as our life spans increase to experience new problems and our ability to detect problems improves, we discover new things that may be killing us or may have been harming us in the past.

That said: I'm unclear on the double standard you were pointing out. Was it something that I said indirectly? If that is the case, the point of my statement is that we have a longer body of evidence for traditional food engineering (selection, cross-breeding, etc) than we do for direct genetic modification by several orders of magnitude--conservatively: 50 years compared to ~5,000 years. That is A) not to say we haven't borked up a few times with traditional engineering and B) not to say that GMO are definitively less safe because they are new. It is just to say that we have definitively less evidence on the matter, and 10-20 years--less than half of a lifetime--is not a resounding endorsement.

All that said: I don't think this is even a particularly significant piece of evidence in the discussion--compared to say: reliable testing standards, risk analysis based upon the changes being introduced rather than the method of introduction, etc--as long as we can agree that 20 years of evident safety does not in itself prove that anything is certainly safe.

[-]Lumifer9y00

My comment was aimed more at one side in the GMO debate rather than specifically at you.

we have a longer body of evidence for traditional food engineering (selection, cross-breeding, etc) than we do for direct genetic modification by several orders of magnitude--conservatively: 50 years compared to ~5,000 years.

This is not true. First, both "traditional food engineering" and GMO are ridiculously broad terms and it's hard to say anything meaningful which applies to the whole category. The main issue, however, is that traditional cross-breeding and such perform major genetic surgery, albeit with crude tools. Look e.g. at this -- you think it's the same corn and wheat? The Green Revolution was so successful precisely because it changed the crops grown. The wheat you're eating is very much not the same thing which was eaten thousands of years ago.

[-]JustinMElms9y00

I think you're drawing two specious conclusions:

First, "traditional food engineering" and GMO do refer to various practices, but there is a very clear distinction of method drawn by those terms. The "traditional" method short circuits natural reproductive process to cultivate desired traits, where as GMO methods entail the direct modification of genes by means external to the reproductive process. To say that repeatedly selecting the largest head of wheat and breeding from that stock is "the same" as injecting new DNA into an organism with a gene gun is absurd in the extreme. They share the same objective, of course, but the method is wholly different.

Second, the Green Revolution was the adoption and expansion of many agricultural practices of which high-yield varieties were one important feature. Obviously, "traditional" methods can have enormous effects. For instance, turning what amounted to an edible grass into a freakish calorie battery. That said, these slow and incremental processes have at least some evolutionary safeguards built into them simply from the time it takes and the holistic, less targeted changes. Once again, we are talking about a difference of method not of objective. The fact that there was a boom of food production prior to GMO does not mean that it is the same as GMO.

[-]Lumifer9y00

here is a very clear distinction of method drawn by those terms

The question is why do you care about methods when you should care only about outcomes.

[-]JustinMElms9y00

You will note that I said that they have the same "objectives" not necessarily the same "outcomes."

Granted, I agree that if we have two genetically and biologically identical organisms, one created by traditional methods and one created by direct genetic modification, then no, I would not care at all.

The argument is that--despite sharing the same objective of improving food production for humanity--traditional methods have a lower likelihood of unforeseen negative outcomes due to the rapid and intricate methods by which GMO are altered.

We care about differences of method because of potential differences of outcome.

[-]Lumifer9y00

The argument is that ... traditional methods have a lower likelihood of unforeseen negative outcomes due to the rapid and intricate methods by which GMO are altered.

That argument doesn't seem persuasive to me. A couple of reasons why: first, I think the "likelihood of unforeseen negative outcomes" in both cases is very vague and uncertain, sufficiently so to make judgement calls about which is lower to be not very credible. Second, in the context of "it's been fine for 20 years but we're not sure about the really long term", I don't see how the "rapid and intricate" quality is relevant.

[-]JustinMElms9y00

I would agree with you that the quoted statement is not terribly persuasive. I was simply encapsulating the actual argument at hand, instead of the straw-man argument of "method versus outcome." And while the vagueness diminishes the magnitude of the evidence, I don't believe it makes it non-zero.

To your second point:

in the context of "it's been fine for 20 years but we're not sure about the really long term", I don't see how the "rapid and intricate" quality is relevant.

I would add to ChristianKI's apt reply that while conventional modifications via breeding can eventually have monumental effects, direct genetic modification can rapidly--over the course of a single generation--have monumental effects that may have unintended side effects attached to them due to a lack of understanding of the intricacies of genetic interactions.

[-]Lumifer9y00

I can't find any sense in ChristianKl's answer, but maybe that's just me.

My basic problem with your position is that "conventional modifications via breeding" are better described as picking from a set of random mutations those where the phenotype looks appealing. I don't know why you think it's a safe method, especially compared with making targeted genetic changes directly.

[-]JustinMElms9y00

I completely agree that breeding methods have their own flaws, which we certainly have seen come to dangerous fruition (pun definitely intended).

I also concede that breeding is quite slow in improving a plant, where direct modification would be much faster.

I furthermore agree that direct genetic modification is the future of crop improvement. Given that we better master the techniques and better understand the genomes in play every year, eventually direct gene modification will lack any of the uncertainty that I invoke right now.

But I likewise think it is not unreasonable to say that it is more likely that we would stumble upon a sudden unfortunate side-effect of our modifications by direct modification, because we would lack the evolutionary "safeguards" that have kept biological life going so far.

In any case: I'm clearly not expressing my ideas cogently enough to be productive in this venue, and it's taken on the whiff of partisan politics. Especially awkward since I am on the same "side" as you: I think there is insufficient evidence to mandate GMO labeling, but I don't like it when "my side" refuses to engage in what I see as reasonable concerns from the "enemy." Once again: not productive.

[-]Lumifer9y00

Yes, I understand we're on the same side. The difference is you think the anti-GMO people have some sort of a case, not quite convincing, but a case. And I think they don't have a case at all and are engaged in spreading pure FUD. I see no reason to search for middle ground with FUD.

[-]ChristianKl9y-10

Second, in the context of "it's been fine for 20 years but we're not sure about the really long term", I don't see how the "rapid and intricate" quality is relevant.

Capabilities of sequencing DNA seems to rise exponentially. Capabilities of making changes via conventional breeding don't. That results in radically different long-term effects.

[+]ChristianKl9y-70

[-]Ishaan9y60

First: check whether the issue is really important: With some exceptions (voting correctly, believing the correct afterlife and not getting sent to hell) If you aren't in a position to interact with the evidence it's probably not something you meaningfully have control over. (Most things for which it is important for you to personally understand have measurable consequences to you. Why do you need the right answer to the GMO question, what would you even do with the right answer?).

Then:

-Figure out exactly what the claims really are and try not to conflate different claims (GMOs will do what, exactly?)

-Consider the possibility that the entire premise is silly ("Is God one or trinity?" "Is she a witch?") and the "consensus" is just wrong and the debate is insane. Generate some plausible third options.

-Check if the two hypotheses seem by your perception to be of roughly equal parsimony, internal logical consistency, and compliance with known evidence, and also check the third options you generated.

-Ask the basic "so, what evidence would you need to tell the difference" questions.

-all the things you mentioned (weigh expert opinions, eliminate bad arguments, eliminate experts who use bad arguments)

-look for concrete predictable things in that area, and adjacent to that area which differ according to the two hypotheses.

-If it's a political issue, try to find out what people who might plausibly be expertish in the area yet don't seem to be invested in debating the issue think about it.

-check what known superforcasters in the field think (people who have a track record of successful predictions in that area). Superforecasters need not actually be loudly engaging with the issue, just ask.

-check if people who have different types of knowledge tend to say different things (e.g. economists vs. sociologists)

-What sorts of knowledge would you need to have to answer the question vs. what sorts of knowledge do the experts in question actually have? (You might think medical doctors are qualified to talk about the effectiveness and safety of various treatments, for example, but they aren't. You want a medical researcher for that. The only difference between a medical doctor and a witch doctor is that one was trained by a curriculum developed by medical researchers and the other wasn't.)

-check for founder effects or cultural effects biasing beliefs (Again, economists vs sociologists. Also, if theologians believe in god at higher rate than biologists it might not be because of different knowledge)

What else? I mean it's a big question, you've asked after a fairly big chunk of "rationality" there.

[-]Tem429y50

Point #2 is a big important point. The media does not select relevant issues, they chose issues that play well to the public. Sometimes these overlap, but often they do not. GMO is a good example, because it is reported as monolithically important, but each genetic modification has to be considered individually; considering GMOs as a unified group is not very useful. Likewise, if you are interested in health and nutrition, you should also look for vegetables that are grown to be nutritious, which includes many GMO but not others: many plants are modified to look better, not be healthier; but many plants are modified to be more nutritious, not look better; etc. Moreover, you can also get some benefit in nutrition by ignoring the GMO debate and looking at things like soil health (organic works as a vague proxy for this, but again, 'organic' is a media chosen label, and so is touted as Very Relevant In Every Way and also does not limit itself to soil health) or time-since-harvest (locally grown, proxy, media pollution, etc.)

[-]JustinMElms9y20

Thanks, Ishaan. That was a lot of good directions to come at this from.

I especially found a few of them novel ways to eke out more confidence from an insulated problem:

If it's a political issue, try to find out what people who might plausibly be expertish in the area yet don't seem to be invested in debating the issue think about it.

check what known superforecasters in the field think (people who have a track record of successful predictions in that area). Superforecasters need not actually be loudly engaging with the issue, just ask.

check if people who have different types of knowledge tend to say different things (e.g. economists vs. sociologists)

I'll try to remember those for questions like this in the future.

Furthermore, notion that you raise struck me:

Most things for which it is important for you to personally understand have measurable consequences to you. Why do you need the right answer to the GMO question, what would you even do with the right answer?

I suppose I've never really considered why I wanted the right answer to a question, I suppose I ascribe a relatively high weight to "understand things" in my utility function. That said, thinking about it from the angle of "What would I do with the right answer": In this case, I would do is embrace/avoid GMO foods for my personal health and safety, vote to label/not-label/ban/regulate GMO, and argue for others to do the same.

Isn't that the ideal of a democratic system: an informed populace vigorously contesting in the marketplace of ideas?

[-]Ishaan9y30

Yes, that is the ideal, and it's true that the three consequences you mention are positive consequences (Assuming more effort makes you more likely to arrive at correct answers, which it usually does although I imagine there are diminishing returns past a certain point - you might notice a lot of very smart people putting a lot of effort into politics and still disagreeing.)

The thing is you must weigh information-gathering and evaluation concerning GMOs against every other possible action you could take with those resources.

Let's focus on the goal which most plausibly requires understanding GMO

for my personal health and safety

Well, let me tell you how i went about researching my personal health and safety:

I researched which foods to eat in general (My conclusions - eat mostly vegetables, meat (try for organ meats and fish), fruits with an overall high fat, low carbohydrate macro-nutrient ratio, avoid vegetable/seed oil, grains. So, in one word, paleolithic. These conclusions are very controversial and I suspect I put in way more effort into researching it than was rationally justified.)

I researched the best way to exercise and learned the techniques (Conclusions: You need to run occasionally and you need to fain flexibility and technique for basic barbell exercises: squat, row, bench, overhead press, etc. I am pretty happy about the time I invested into researching these.)

I've put moderate effort into researching basic pesticide avoidance (there are lists of highest pesticide foods you can avoid buying), ethical meat sourcing, and ecologically sustainable fish sourcing. Ultimately I've put very little effort into this relative to the first two.)

I've skimmed examine.com for potentially helpful supplements (Conclusions: Fish oil, Vit D, Vit K2magnesium (ZMA, don't use MgO it's not bioavailable. I probably spent too much time on this.)

GMOs are pretty far down on this list of things which I think are probably important. I haven't really gotten to them yet.

Do you see where the prioritization issue comes in here? And that's when your personal health is the main goal. The chance that GMO is high on the priority list in the genre of public dietary health, is in my mind, pretty minuscule. If you narrow your specialization to "regulatory mechanisms concerning food", then it'll be worth studying GMOs as one of the branches in your knowledge tree, but probably not before you've studied broad stuff about regulatory mechanisms first. (as I understand it, GMOs are not a monolithic thing so it's more interesting to study start with general stuff about how innovations in food are handled, etc).

You don't necessarily need to agree with me about prioritization, but you should spend some time thinking about prioritization.

I suppose I ascribe a relatively high weight to "understand things" in my utility function

Of course, we all do. But there is a whole world of things, so, which things, and why? Information due to purely Intrinsic interest is malleable

[-]ChristianKl9y50

Apart from whether or not GMO's are healthy we general believe that customers should get information about products that they want to have. We don't allow artificial diamonds to be sold as naturally mined diamonds, even when De Beers doesn't have an argument that artifical diamonds are unsafe.

[-]Viliam9y00

I generally agree with your position, but let me be the devil's advocate here:

Let's assume that some fraction of the population are Nazis, and they decide to boycott all Jewish products. Let's assume there is a general agreement on the definition of "Jewish products" (for example anything produced by a company with a Jewish owner, or using such component). The Nazis express a desire to have all products clearly labeled. Would you support this specific customer right?

[-]Lumifer9y20

Heh.

[-]ChristianKl9y00

The idea that products are labeled with the country of origin are quite standard. It's equivalent to how products made in China can't be sold as "made in the US".

[-]Viliam9y00

products made in China can't be sold as "made in the US"

Is there a standard simple way to circumvent this rule? Such as establishing a one-employee branch of your company in USA and pretending that the employee did some important finishing touch on your product, before you sell it on the American market?

[-]Lumifer9y00

If you were to read the link carefully, you would notice that the regulations in question prohibit the "Made in Israel" label :-/

[-]ChristianKl9y00

For products not made in Israel but in territory that isn't in Israel according to the UN decisions about what happens to be Israel.

[-]Lumifer9y00

And now go and look at OP again. The EU directive insists on having labels "product from the West Bank (Israeli settlement)" and "product from the West Bank (Palestinian product)".

[-]ChristianKl9y00

The core issue of the conflict is whether "made in Israel" can be written on products that aren't made in the official territory of Israel as set by the UN.

I don't think that produces of products labeled "product from the West Bank (Israeli settlement)" want to write "produced in Palestina" on their products, so I see no problem with offering them that label. Furthermore those products simply might not be covered by the trade deals with the Palestianian authority and therefore there might be a need to distinguish them from other West Bank products.

[-]ChristianKl9y00

There are clearly layed out categories that by law can't be used to discriminate people. Religion is among them. I don't think that information about protect classes has to be provided.

Moderation Log