= 27d567bc2d48d9f40e9ccc20ea370b15)
Elo comments on Rationality when Insulated from Evidence - Less Wrong Discussion
You are viewing a comment permalink. View the original post to see all comments and the full post content.
= 783df68a0f980790206b9ea87794c5b6)
Loading…= b715d02e85f7b3b4ae65ca3d3e450868)
Subscribe to RSS Feed
Powered by Reddit
= f037147d6e6c911a85753b9abdedda8d)
You are viewing a comment permalink. View the original post to see all comments and the full post content.
Comments (60)
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.
Just like the turkey had a lot of evidence for humans being nice to him until the day before thanksgiving.
By the standard that 20 years with a new technology should be enough to see problems with it various techonologies from lead pipes, to cigaretts to asbest, were also proven to be safe.
Without labeling of products it's also difficult to actually gather the information. I think it's a bad general argument to say that people shouldn't know whether they are ingesting X because X isn't proven to do anything yet.
That's true, 20 years wouldn't necessarily bring to light a delayed effect.
However the GMO case is interesting because we have in effect a massive scale natural experiment, where hundreds of millions of people on one continent have eaten lots of GMO food while hundreds of millions on another continent have eaten very little, over a period of 10-15 years. There is also a highly motivated group of people who bring to the public attention even the smallest evidence of harm from GMOs.
While I don't rule out a harmful long-term effect, GMOs are a long way down on my list of things to worry about, and dropping further over time.
Not really, because the two groups differs in many attributes. You can't draw any reliable conclusions from that if you don't know individual consumption. If you could draw that conclusion we could conclude from US bee deaths that GMO's are bad.
But there also no reason to assume that risk from GMO would be equally distributed among different GMO foods. Letting plants produce poisons so that they won't get eaten by insects is likely more risky than doing something to improve drought resistance.
Our ability to manipulate organism increases as time goes on. Organisms where multiple genes are added might be more risky than organisms where only a single gene was added.
Valid arguments against early GMO that they spread antibiotic resistance genes also don't hold against newer GMO's.
Bioengineered pandemics frequently top the LW census as an X-risk concern. Commerical usage of GMO's pays for technology development to produce more capabilities on that front.
yes.
One difficulty about GMO specifically is that as you said,
Golden rice - probably fine. Pesticide resistant stuff, probably not as fine for various reasons already published in the public domain.
The problem is when talking about GMO you cover the existing proven bad (and since no longer used) as well as the unproven bad. As well as the good and the unproven good and the neutral. you just talk about GMO. It might help to be more specific. Can you be more specific about what you are calling GMO? And what you are saying is the problem? There is no inherent problem with the concept of GMO (modifying genes). It depends on how you use it (which genes and how you modify).
In a world of labeling I have no problem with having more specific labels for different types of it.
Goodhard's law is generally a problem when you have strong optimisation tools.
With unlabeled GMO's the commercial pressure is to create food that is as cheap as possible without regard for whether it's healthy. If you require labeling than the companies producing the food have incentives to produce healthy food.
GMO's reduce diversity of agriculture. That produces a systems that generally less robust, for reasons that Nassim Taleb talks about frequently.
Do you believe that people shouldn't know whether or not their rice has added Vitamin A? I think it's very worthwhile for people to know about it.
You are jumping topic. GMO risk is different from GMO labeling. However, it is true that labeling nutrition information is good, regardless of GMO status, and that GMO may have more variation in nutritional content (positive and negative) than non-GMO.
No, there are practical healths risks that come from food containing substances that I don't expect. I might get too much of a certain vitamin if I don't know that it's added to my food.
Yes... but this is not an issue of GMO. This is an issue of additives. You should require information that is clearly relevant to health regardless of GMO status. GMOing is a way of adding nutrients, but we would want additives labeled regardless of how they are added.
Or, to put it another way, this is a case where the GMO change is something that should be labeled, because there is a possible effect on health. But the factor under consideration isn't that it is GMO, it is that it there is an possible effect on health.
No, we don't want to require everybody who sells an orange to pay for lab tests that determine for every vitamin how much is contained. Making such a requirement would be a death sentence for farmers markets.
A customers has certain expecations about what an natural orange happens to be. It's a class of objects that shares basic traits. GMO allows giving the orange traits that oranges generally aren't expected to have.
Any GMO interventions adds new molecules. If you follow that framework, if you add genes that produce 3 new proteins, put those three proteins on the label.
As another example of probably fine GMO, I've just come across a review highlighting the urgency of engineering microbes able to assimilate lactose (to use whey and other wastes of dairy industry as substrates for exopolysaccharide synthesis). They also argue for creating (more) efficient EPS-producers culturable on cellulose-containing wastes, although that does seem to me rather more dangerous, on technical glycerole after biodiesel production, etc.
I think I disagree on these examples. I don't know that these were all proven safe. And not for very long. Even when they were proven "safe" temporarily, there were some science or medical events that caused concern.
Your idea of "proven-safe" is that they were used for two decades without proof of problems. Those examples fit that standard.
Were they used for two decades without strong evidence of problems while people were looking for such evidence?
They look to me like things that no one had any idea were problematic -- until at some point evidence of trouble started to appear, and fairly quickly it became consensus that they were bad. (Possible counterexample -- I'm not sure of the dates -- is cigarettes, only because there was a very well-funded systematic disinformation campaign conducted by the tobacco companies.)
In the case of GMO, some people have worried (publicly, vocally) about safety from the very beginning. That's quite different from lead pipes or asbestos or even cigarettes.
The standard should be comparing outcomes in users and non-users after 20 years. The idea is that 20 years is enough to show any effect that exists, and we shouldn't refrain from adopting the new thing waiting for even more evidence to come in.
Cigarettes, asbestos and leaded water would all show strong effects after 20 years, not as strong as after 50 years, but certainly enough to identify a problem.
Certainly some things have a very delayed effect, but they are very very few compared to things that have a quick effect; most foods that are bad for you show an effect within hours to months. We shouldn't treat every possible new food as having a significant risk of an effect 30 years later unless there's a specific reason (plausible mechanism).
Why? Especially when the discussion is not about banning the food but about people's right to know that they are eating a new food.
But even if that's true it's besides the point because GMOs aren't "a food" but a group of a large amount of different foods.
Letting plants produce poisons to not get eaten by insects suggest to me a plausible mechanism that involves the poison also harming humans.
Because if we're too suspicious, we pay the opportunity cost of whatever makes it an attractive new food in the first place.
The problem is that this bakes in certain assumptions about what makes a food new in a potentially dangerous way and so requires mandatory labeling.
Agricultural technology is always changing. We don't require labeling for most of the changes, even though our prior for their potential danger might be much higher than for GMOs. Examples of things that don't require labeling: which pesticides and antibiotics were used, and in what amounts; what diseases and parasites may have been present; what the storage and transportation conditions were.
For all of these things there are regulatory frameworks. But I can't think offhand of any examples, other than GMOs, where the legal status is "you may do X, but you have to label it appropriately". E.g., there's no "you may use the new pesticide X, but you have to label each piece of fruit sold as being X-positive". When people want to signal they're not using something, it's up to them to label produce as "organic" or "X-free"; everyone else doesn't have to label theirs as "non-organic". This difference sends a strong signal to the public that GMOs are presumed to be more dangerous (or risky/unproven) than every other legal agriculture technology.
And, as you point out, generalizing and regulating all GMOs as a group makes no more sense than regulating all pesticides as a group.
I wouldn't have a problem if we distinguish GMOs into different classes and put the resulting class on the label.
That's not really true. Ingredient lists require the disclose of many substances that are added to new foods. People have a right to know whether their food contains aspartame.
I would also support requiring big producers to provide that information. Products could have a barcode that can be scanned and the information could be provided via the internet.
Certainly, I agree: there is no reason that we shouldn't be able to know every detail about the materials and processes that go into our food, but surely you acknowledge the connotative difference between:
"Scan this to see all relevant information"
and
"Governmental authority mandates that we declare this food to contain GMO"
There are issues with compliance costs that make it hard to force disclosure of all information. The compliance costs fo writing GMO wheat instead of wheat on a ingridient box are little.
I would be happy if the companies would have a choice to put up a scan code that provides all relevant information in exchange for not having to write things on the label.
This is a good point. But it still stands in contrast to non-disclosure of everything that's not an ingridient: processes, pesticides, etc. Produce like fruit or raw meat doesn't have any "ingridiends".
Why the difference? I lean towards thinking it's in large part historical, political, and accidental, rather than reflecting any real difference in what's appropriate or required.
GMO's are ingridients.
Golden Rice looks different than normal rice, so people who want to buy it can see the difference and make informed decisions about what they want to buy. With a lot of other GMO products that isn't the case.
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.
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:
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".
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?
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.
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.
My comment was aimed more at one side in the GMO debate rather than specifically at you.
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.
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.
The question is why do you care about methods when you should care only about outcomes.
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.