Settling Titan, Schneier’s Law, and scenario thinking

Charles Wohlforth and Amanda R. Hendrix want us to colonize Titan. The essay irritated me in an interesting manner.

Full disclosure: they interviewed me while they were writing their book Beyond Earth: Our Path to a New Home in the Planets, which I have not read yet, and I will only be basing the following on the SciAm essay. It is not really about settling Titan either, but something that bothers me with a lot of scenario-making.

A weak case for Titan and against Luna and Mars

Basically the essay outlines reasons why other locations in the solar system are not good: Mercury too hot, Venus way too hot, Mars and Luna have too much radiation. Only Titan remains, with a cold environment but not too much radiation.

A lot of course hinges on the assumptions:

We expect human nature to stay the same. Human beings of the future will have the same drives and needs we have now. Practically speaking, their home must have abundant energy, livable temperatures and protection from the rigors of space, including cosmic radiation, which new research suggests is unavoidably dangerous for biological beings like us.

I am not that confident in that we will remain biological or vulnerable to radiation. But even if we decide to accept the assumptions, the case against the Moon and Mars is odd:

Practically, a Moon or Mars settlement would have to be built underground to be safe from this radiation.Underground shelter is hard to build and not flexible or easy to expand. Settlers would need enormous excavations for room to supply all their needs for food, manufacturing and daily life.

So making underground shelters is much harder than settling Titan, where buildings need to be isolated against a -179 C atmosphere and ice ground full with complex and quite likely toxic hydrocarbons. They suggest that there is no point in going to the moon to live in an underground shelter when you can do it on Earth, which is not too unreasonable – but is there a point in going to live inside an insulated environment on Titan either? The actual motivations would likely be less of a desire for outdoor activities and more scientific exploration, reducing existential risk, and maybe industrialization.

Also, while making underground shelters in space may be hard, it does not look like an insurmountable problem. The whole concern is a bit like saying submarines are not practical because the cold of the depths of the ocean will give the crew hypothermia – true, unless you add heating.

I think this is similar to Schneier’s law:

Anyone, from the most clueless amateur to the best cryptographer, can create an algorithm that he himself can’t break.

It is not hard to find a major problem with a possible plan that you cannot see a reasonable way around. That doesn’t mean there isn’t one.

Settling for scenarios

Maybe Wohlforth and Hendrix spent a lot of time thinking about lunar excavation issues and consistent motivations for settlements to reach a really solid conclusion, but I suspect that they came to the conclusion relatively lightly. It produces an interesting scenario: Titan is not the standard target when we discuss where humanity ought to go, and it is an awesome environment.

Similarly the “humans will be humans” scenario assumptions were presumably chosen not after a careful analysis of relative likelihood of biological and postbiological futures, but just because it is similar to the past and makes an interesting scenario. Plus human readers like reading about humans rather than robots. All together it makes for a good book.

Clearly I have different priors compared to them on the ease and rationality of Lunar/Martian excavation and postbiology. Or even giving us D. radiodurans genes.

In The Age of Em Robin Hanson argues that if we get the brain emulation scenario space settlement will be delayed until things get really weird: while postbiological astronauts are very adaptable, so much of the mainstream of civilization will be turning inward towards a few dense centers (for economics and communications reasons). Eventually resource demand, curiosity or just whatever comes after the Age of Ems may lead to settling the solar system. But that process will be pretty different even if it is done by mentally human-like beings that do need energy and protection. Their ideal environments would be energy-gradient rich, with short communications lags: Mercury, slowly getting disassembled into a hot Dyson shell, might be ideal. So here the story will be no settlement, and then wildly exotic settlement that doesn’t care much about the scenery.

But even with biological humans we can imagine radically different space settlement scenarios, such as the Gerhard O’Neill scenario where planetary surfaces are largely sidestepped for asteroids and space habitats. This is Jeff Bezo’s vision rather than Elon Musk’s and Wohlforth/Hendrix’s. It also doesn’t tell the same kind of story: here our new home is not in the planets but between them.

My gripe is not against settling Titan, or even thinking it is the best target because of some reasons. It is against settling too easily for nice scenarios.

Beyond the good story

Sometimes we settle for scenarios because they tell a good story. Sometimes because they are amenable to study among other, much less analyzable possibilities. But ideally we should aim at scenarios that inform us in a useful way about options and pathways we have.

That includes making assumptions wide enough to cover relevant options, even the less glamorous or tractable ones.

That requires assuming future people will be just as capable (or more) at solving problems: just because I can’t see a solution to X doesn’t mean it is not trivially solved in the future.

(Maybe we could call it the “Manure Principle” after the canonical example of horse manure being seen as a insoluble urban planning problem at the previous turn of century and then neatly getting resolved by unpredicted trams and cars – and just like Schneier’s law and Stigler’s law the reality is of course more complex than the story.)

In standard scenario literature there are often admonitions not just to select a “best case scenario”, “worst case scenario” and “business as usual scenario” – scenario planning comes into its own when you see nontrivial, mixed value possibilities. In particular, we want decision-relevant scenarios that make us change what we will do when we hear about them (rather than good stories, which entertain but do not change our actions). But scenarios on their own do not tell us how to make these decisions: they need to be built from our rationality and decision theory applied to their contents. Easy scenarios make it trivial to choose (cake or death?), but those choices would have been obvious even without the scenarios: no forethought needed except to bring up the question. Complex scenarios force us to think in new ways about relevant trade-offs.

The likelihood of complex scenarios is of course lower than simple scenarios (the conjunction fallacy makes us believe much more in rich stories). But if they are seen as tools for developing decisions rather than information about the future, then their individual probability is less of an issue.

In the end, good stories are lovely and worth having, but for thinking and deciding carefully we should not settle for just good stories or the scenarios that feel neat.

Solomon’s frozen judgement

A girl dying of cancer wanted to use cryonic preservation to have a chance at being revived in the future. While supported by her mother the father disagreed; in a recent high court ruling, the judge found that she could be cryopreserved.

As the judge noted, the verdict was not a statement on the validity of cryonics itself, but about how to make decisions about prospective orders. In many ways the case would presumably have gone the same way if there had been a disagreement about whether the daughter could have catholic last rites. However, cryonics makes things fresh and exciting (I have been in the media all day thanks to this).

What is the ethics of parents disagreeing about the cryosuspension of their child?

Best interests

One obvious principle is that parents ought to act in the best interest of their children.

If the child is morally mature and with informed consent, then they can clearly have a valid interest in taking a chance on cryonics: they might not be legally adult, but as in normal medical ethics their stated interests have strong weight. Conversely, one could imagine a case where a child would not want to be preserved, in which case I think most people would agree their preferences should dominate.

The general legal consensus in the West is that the child’s welfare is so important that it can overrule the objections of parents. In UK law parents have the right and the duty to give consent for a minor. Children can consent for medical treatment, overriding their parents, at 16. However, if refusing treatment parents and court can override. This mostly comes into play in cases such as avoiding blood transfusions for religious reasons.

In this case the issue was that the parents were disagreeing and the child was not legally old enough.

If one thinks cryonics is reasonable, then one should clearly cryosuspend the child: it is in their best interest. But if one thinks cryonics is not reasonable, is it harming the interest of the child? This seems to require some theory of how cryonics is bad for the interests of the child.

As an analogy, imagine a case where one parent is a Jehovah’s Witness and want to refuse a treatment involving blood transfusion: the child will die without the treatment, and it will be a close call even with it. Here the objecting parent may claim that undergoing the transfusion harms the child in an important spiritual way and refuse consent. The other parent disagrees. Here the law would come down on the side of the pro-transfusion parent.

On this account and if we agree the cases are similar, we might say that parents have a legal duty to consent to cryonics.

Weak and strong reasons

In practice the controversialness of cryonics may speak against this: many people disagree about cryonics being good for one’s welfare. However, most such arguments usually seem to be based on various farfetched scenarios about how the future could be a bad place to end up in. Others bring up loss of social connections or that personal identity would be disrupted. A more rational argument is that it is an unproven treatment of dubious efficacy, which would make it irrational to undertake if there was an alternative; however since there isn’t any alternative this argument has little power. The same goes for the risk of loss of social connection or identity: had there been an alternative to death (which definitely severs connections and dissolves identity) that may have been preferable. If one seriously thinks that the future would be so dark that it is better not to get there, one should probably not have children.

In practice it is likely that the status of cryonics as nonstandard treatment would make the law hesitate to overrule parents. We know blood transfusions work, and while spiritual badness might be a respectable as a private view we as a society do not accept it as a sufficient reason to have somebody die. But in the case of cryonics the unprovenness of the treatment means that hope for revival is on nearly the same epistemic level as spiritual badness: a respectable private view, but not strong enough to be a valid public reason. Cryonicists are doing their best to produce scientific evidence – tissue scans, memory experiments, protocols – that move the reasons to believe in cryonics from the personal faith level to the public evidence level. They already have some relevant evidence. As soon as lab mice are revived or people become convinced the process saves the connectome the reasons would be strengthened and cryonics becomes more akin blood transfusion.

The key difference is that weak private reasons are enough to allow an experimental treatment where there is no alternative but death, but they are generally not enough to go for an experimental treatment when there is some better treatment. When disallowing a treatment weak reasons may work well against unproven or uncertain treatments, but not when it is proven. However, disallowing a treatment with no alternative is equivalent to selecting death.

When two parents disagree about cryonics (and the child does not have a voice) it hence seems that they both have weak reasons, but the asymmetry between having a chance and dying tilts in favor of cryonics. If it was purely a matter of aesthetics or value (for example, arguing about the right kind of last rites) there would be no societal or ethical constraint. But here there is some public evidence, making it at least possible that the interests of the child might be served by cryonics. Better safe than sorry.

When the child also has a voice and can express its desires, then it becomes obvious which way to go.

King Solomon might have solved the question by cryosuspending the child straight away, promising the dissenting parent not to allow revival until they either changed their mind or there was enough public evidence to convince anybody that it would be in the child’s interest to be revived. The nicest thing about cryonics is that it buys you time to think things through.

AI, morality, ethics and metaethics

Next Sunday I will be debating AI ethics at Battle of Ideas. Here is a podcast where I talk AI, morality and ethics: https://soundcloud.com/institute-of-ideas/battle-cry-anders-sandberg-on-ethical-ai

What distinguishes morals from ethics?

There is actually a shocking confusion about what the distinction between morals and ethics is. Differen.com says ethics is about rules of conduct produced by an external source while morals are an individual’s own principles of right and wrong. Grammarist.com says morals are principles on which one’s own judgement of right and wrong are based (abstract, subjective and personal), ethics are the principles of right conduct (practical, social and objective). Ian Welsh gives a soundbite: “morals are how you treat people you know.  Ethics are how you treat people you don’t know.” Paul Walker and Terry Lovat say ethics leans towards decisions based on individual character and subjective understanding of right and wrong, while morals is about widely shared communal or societal norms – here ethics is individual assessment of something being good or bad, while morality is inter-subjective community assessment.

Wikipedia distinguishes between ethics as a research field and the common human ability to think critically about moral values and direct actions appropriately, or a particular persons principles of values. Morality is the differentiation between things that are proper and improper, as well as a body of standards and principles in derived from a code of conduct in some philosophy, religion or culture… or derived from a standard a person believes to be universal.

Dictionary.com regards ethics as a system of moral principles, the rules of conduct recognized in some human environment, an individual’s moral principles (and the branch of philosophy). Morality is about conforming to the rules of right conduct, having moral quality or character, a doctrine or system of morals and a few other meanings. The Cambridge dictionary thinks ethics is the study of what is right or wrong, or the set of beliefs about it, while morality is a set of personal or social standards for good/bad behavior and character.

And so on.

I think most people try to include the distinction between shared systems of conduct and individual codes, and the distinction between things that are subjective, socially agreed on, and maybe objective. Plus that we all agree on that ethics is a philosophical research field.

My take on it

I like to think of it as a AI issue. We have a policy function $\pi(s,a)$ that maps states and action pairs to a probability of acting that way; this is set using a value function $Q(s)$ where various states are assigned values. Morality in my sense is just the policy function and maybe the value function: they have been learned through interacting with the world in various ways.

Ethics in my sense is ways of selecting policies and values. We are able to not only change how we act but also how we evaluate things, and the information that does this change is not just reward signals that update value function directly, but also knowledge about the world, discoveries about ourselves, and interactions with others – in particular ideas that directly change the policy and value functions.

When I realize that lying rarely produces good outcomes (too much work) and hence reduce my lying, then I am doing ethics (similarly, I might be convinced about this by hearing others explain that lying is morally worse than I thought or convincing me about Kantian ethics). I might even learn that short-term pleasure is less valuable than other forms of pleasure, changing how I view sensory rewards.

Academic ethics is all about the kinds of reasons and patterns we should use to update our policies and values, trying to systematize them. It shades over into metaethics, which is trying to understand what ethics is really about (and what metaethics is about: it is its own meta-discipline, unlike metaphysics that has metametaphysics, which I think is its own meta-discipline).

I do not think I will resolve any confusion, but at least this is how I tend to use the terminology. Morals is how I act and evaluate, ethics is how I update how I act and evaluate, metaethics is how I try to think about my ethics.

How much should we spread out across future scenarios?

Robin Hanson mentions that some people take him to task for working on one scenario (WBE) that might not be the most likely future scenario (“standard AI”); he responds by noting that there are perhaps 100 times more people working on standard AI than WBE scenarios, yet the probability of AI is likely not a hundred times higher than WBE. He also notes that there is a tendency for thinkers to clump onto a few popular scenarios or issues. However:

In addition, due to diminishing returns, intellectual attention to future scenarios should probably be spread out more evenly than are probabilities. The first efforts to study each scenario can pick the low hanging fruit to make faster progress. In contrast, after many have worked on a scenario for a while there is less value to be gained from the next marginal effort on that scenario.

This is very similar to my own thinking about research effort. Should we focus on things that are likely to pan out, or explore a lot of possibilities just in case one of the less obvious cases happens? Given that early progress is quick and easy, we can often get a noticeable fraction of whatever utility the topic has by just a quick dip. The effective altruist heuristic of looking at neglected fields also is based on this intuition.

A model

But under what conditions does this actually work? Here is a simple model:

There are $N$ possible scenarios, one of which ($j$) will come about. They have probability $P_i$. We allocate a unit budget of effort to the scenarios: $\sum a_i = 1$. For the scenario that comes about, we get utility $\sqrt{a_j}$ (diminishing returns).

Here is what happens if we allocate proportional to a power of the scenarios, $a_i \propto P_i^\alpha$. $\alpha=0$ corresponds to even allocation, 1 proportional to the likelihood, >1 to favoring the most likely scenarios. In the following I will run Monte Carlo simulations where the probabilities are randomly generated each instantiation. The outer bluish envelope represents the 95% of the outcomes, the inner ranges from the lower to the upper quartile of the utility gained, and the red line is the expected utility.

This is the $N=2$ case: we have two possible scenarios with probability $p$ and $1-p$ (where $p$ is uniformly distributed in [0,1]). Just allocating evenly gives us $1/\sqrt{2}$ utility on average, but if we put in more effort on the more likely case we will get up to 0.8 utility. As we focus more and more on the likely case there is a corresponding increase in variance, since we may guess wrong and lose out. But 75% of the time we will do better than if we just allocated evenly. Still, allocating nearly everything to the most likely case means that one does lose out on a bit of hedging, so the expected utility declines slowly for large $\alpha$.

The  $N=100$ case (where the probabilities are allocated based on a flat Dirichlet distribution) behaves similarly, but the expected utility is smaller since it is less likely that we will hit the right scenario.

What is going on?

This doesn’t seem to fit Robin’s or my intuitions at all! The best we can say about uniform allocation is that it doesn’t produce much regret: whatever happens, we will have made some allocation to the possibility. For large N this actually works out better than the directed allocation for a sizable fraction of realizations, but on average we get less utility than betting on the likely choices.

The problem with the model is of course that we actually know the probabilities before making the allocation. In reality, we do not know the likelihood of AI, WBE or alien invasions. We have some information, and we do have priors (like Robin’s view that $P_{AI} < 100 P_{WBE}$), but we are not able to allocate perfectly.  A more plausible model would give us probability estimates instead of the actual probabilities.

We know nothing

Let us start by looking at the worst possible case: we do not know what the true probabilities are at all. We can draw estimates from the same distribution – it is just that they are uncorrelated with the true situation, so they are just noise.

In this case uniform distribution of effort is optimal. Not only does it avoid regret, it has a higher expected utility than trying to focus on a few scenarios ($\alpha>0$). The larger N is, the less likely it is that we focus on the right scenario since we know nothing. The rationality of ignoring irrelevant information is pretty obvious.

Note that if we have to allocate a minimum effort to each investigated scenario we will be forced to effectively increase our $\alpha$ above 0. The above result gives the somewhat optimistic conclusion that the loss of utility compared to an even spread is rather mild: in the uniform case we have a pretty low amount of effort allocated to the winning scenario, so the low chance of being right in the nonuniform case is being balanced by having a slightly higher effort allocation on the selected scenarios. For high $\alpha$ there is a tail of rare big “wins” when we hit the right scenario that drags the expected utility upwards, even though in most realizations we bet on the wrong case. This is very much the hedgehog predictor story: ocasionally they have analysed the scenario that comes about in great detail and get intensely lauded, despite looking at the wrong things most of the time.

We know a bit

We can imagine that knowing more should allow us to gradually interpolate between the different results: the more you know, the more you should focus on the likely scenarios.

If we take the mean of the true probabilities with some randomly drawn probabilities (the “half random” case) the curve looks quite similar to the case where we actually know the probabilities: we get a maximum for $\alpha\approx 2$. In fact, we can mix in just a bit ($\beta$) of the true probability and get a fairly good guess where to allocate effort (i.e. we allocate effort as $a_i \propto (\beta P_i + (1-\beta)Q_i)^\alpha$ where $Q_i$ is uncorrelated noise probabilities). The optimal alpha grows roughly linearly with $\beta$, $\alpha_{opt} \approx 4\beta$ in this case.

We learn

Adding a bit of realism, we can consider a learning process: after allocating some effort $\gamma$ to the different scenarios we get better information about the probabilities, and can now reallocate. A simple model may be that the standard deviation of noise behaves as $1/\sqrt{\tilde{a}_i}$ where $\tilde{a}_i$ is the effort placed in exploring the probability of scenario $i$. So if we begin by allocating uniformly we will have noise at reallocation of the order of $1/\sqrt{\gamma/N}$. We can set $\beta(\gamma)=\sqrt{\gamma/N}/C$, where $C$ is some constant denoting how tough it is to get information. Putting this together with the above result we get $\alpha_{opt}(\gamma)=\sqrt{2\gamma/NC^2}$. After this exploration, now we use the remaining $1-\gamma$ effort to work on the actual scenarios.

This is surprisingly inefficient. The reason is that the expected utility declines as $\sqrt{1-\gamma}$ and the gain is just the utility difference between the uniform case $\alpha=0$ and optimal $\alpha_{opt}$, which we know is pretty small. If C is small (i.e. a small amount of effort is enough to figure out the scenario probabilities) there is an optimal nonzero  $\gamma$. This optimum $\gamma$ decreases as C becomes smaller. If C is large, then the best approach is just to spread efforts evenly.

Conclusions

So, how should we focus? These results suggest that the key issue is knowing how little we know compared to what can be known, and how much effort it would take to know significantly more.

If there is little more that can be discovered about what scenarios are likely, because our state of knowledge is pretty good, the world is very random,  or improving knowledge about what will happen will be costly, then we should roll with it and distribute effort either among likely scenarios (when we know them) or spread efforts widely (when we are in ignorance).

If we can acquire significant information about the probabilities of scenarios, then we should do it – but not overdo it. If it is very easy to get information we need to just expend some modest effort and then use the rest to flesh out our scenarios. If it is doable but costly, then we may spend a fair bit of our budget on it. But if it is hard, it is better to go directly on the object level scenario analysis as above. We should not expect the improvement to be enormous.

Here I have used a square root diminishing return model. That drives some of the flatness of the optima: had I used a logarithm function things would have been even flatter, while if the returns diminish more mildly the gains of optimal effort allocation would have been more noticeable. Clearly, understanding the diminishing returns, number of alternatives, and cost of learning probabilities better matters for setting your strategy.

In the case of future studies we know the number of scenarios are very large. We know that the returns to forecasting efforts are strongly diminishing for most kinds of forecasts. We know that extra efforts in reducing uncertainty about scenario probabilities in e.g. climate models also have strongly diminishing returns. Together this suggests that Robin is right, and it is rational to stop clustering too hard on favorite scenarios. Insofar we learn something useful from considering scenarios we should explore as many as feasible.

The case for Mars

On practical Ethics I post about the goodness of being multi-planetary: is it rational to try to settle Mars as a hedge against existential risk?

The problem is not that it is absurd to care about existential risks or the far future (which was the Economist‘s unfortunate claim), nor that it is morally wrong to have a separate colony, but that there might be better risk reduction strategies with more bang for the buck.

One interesting aspect is that making space more accessible makes space refuges a better option. At some point in the future, even if space refuges are currently not the best choice, they may well become that. There are of course other reasons to do this too (science, business, even technological art).

So while existential risk mitigation right now might rationally aim at putting out the current brushfires and trying to set the long-term strategy right, doing the groundwork for eventual space colonisation seems to be rational.

Doing right and feeling good

My view is largely that moral action is strongly driven and motivated by emotions rather than reason, but outside the world of the blindingly obvious or everyday human activity our intuitions and feelings are not great guides. We do not function well morally when the numbers get too big or the cognitive biases become maladaptive. Morality may be about the heart, but ethics is in the brain.

What makes a watchable watchlist?

Stefan Heck managed to troll a lot of people into googling “how to join ISIS”. Very amusing, and now a lot of people think they are on a NSA watchlist.

This kind of prank is of course by why naive keyword-based watch lists are total failures. One prank and it gets overloaded. I would be shocked if any serious intelligence agency actually used them for real. Given that people’s Facebook likes give pretty good predictions of who they are (indeed, better than many friends know them) there are better methods if you happen to be a big intelligence agency.

Still, while text and other online behavior signal a lot about a person, it might not be a great tool for making proper watchlists since there is a lot of noise. For example, this paper extracts personality dimensions from online texts and looks at civilian mass murderers. They state:

Using this ranking procedure, it was found that all of the murderers’ texts were located within the highest ranked 33 places. It means that using only two simple measures for screening these texts, we can reduce the size of the population under inquiry to 0.013% of its original size, in order to manually identify all of the murderers’ texts.

At first, this sounds great. But for the US, that means the watchlist for being a mass murderer would currently have 41,000 entries. Given that over the past 150 years there has been about 150 mass murders in the US, this suggests that the precision is not going to be that great – most of those people are just normal people. The base rate problem crops up again and again when trying to find rare, scary people.

The deep problem is that there is not enough positive data points (the above paper used seven people) to make a reliable algorithm. The same issue cropped up with NSA’s SKYNET program – they also had seven positive examples and hundreds of thousands of negatives, and hence had massive overfitting (suggesting the Islamabad Al Jazeera bureau chief was a prime Al Qaeda suspect).

Rational watchlists

The rare positive data point problem strikes any method, no matter what it is based on. Yes, looking at the social network around people might give useful information, but if you only have a few examples of bad people the system will now pick up on networks like the ones they had. This is also true for human learning: if you look too much for people like the ones that in the past committed attacks, you will focus too much on people like them and not enemies that look different. I was told by an anti-terrorism expert about a particular sign for veterans of Afghan guerrilla warfare: great if and only if such veterans are the enemy, but rather useless if the enemy can recruit others. Even if such veterans are a sizable fraction of the enemy the base rate problem may make you spend your resources on innocent “noise” veterans if the enemy is a small group. Add confirmation bias, and trouble will follow.

Note that actually looking for a small set of people on the watchlist gets around the positive data point problem: the system can look for them and just them, and this can be made precise. The problem is not watching, but predicting who else should be watched.

The point of a watchlist is that it represents a subset of something (whether people or stocks) that merits closer scrutiny. It should essentially be an allocation of attention towards items that need higher level analysis or decision-making. The U.S. Government’s Consolidated Terrorist Watch List requires nomination from various agencies, who presumably decide based on reasonable criteria (modulo confirmation bias and mistakes). The key problem is that attention is a limited resource, so adding extra items has a cost: less attention can be spent on the rest.

This is why automatic watchlist generation is likely to be a bad idea, despite much research. Mining intelligence to help an analyst figure out if somebody might fit a profile or merit further scrutiny is likely more doable. As long as analyst time is expensive it can easily be overwhelmed if something fills the input folder: HUMINT is less likely to do it than SIGINT, even if the analyst is just doing the preliminary nomination for a watchlist.

The optimal Bayesian watchlist

One can analyse this in a Bayesian framework: assume each item has a value $x_i$ distributed as $f(x_i)$. The goal of the watchlist is to spend expensive investigatory resources to figure out the true values; say the cost is 1 per item. Then a watchlist of randomly selected items will have a mean value $V=E[x]-1$. Suppose a cursory investigation costing much less gives some indication about $x_i$, so that it is now known with some error: $y_i = x_i+\epsilon$. One approach is to select all items above a threshold $\theta$, making $V=E[x_i|y_i<\theta]-1$.

If we imagine that everything is Gaussian $x_i \sim N(\mu_x,\sigma_x^2), \epsilon \sim N(0,\sigma_\epsilon^2)$, then  $V=\int_\theta^\infty t \phi(\frac{t-\mu_x}{\sigma_x}) \Phi\left(\frac{t-\mu_x}{\sqrt{\sigma_x^2+\sigma_\epsilon^2}}\right)dt$. While one can ram through this using Owen’s useful work, here is a Monte Carlo simulation of what happens when we use $\mu_x=0, \sigma_x^2=1, \sigma_\epsilon^2=1$ (the correlation between x and y is 0.707, so this is not too much noise):

Note that in this case the addition of noise forces a far higher threshold than without noise (1.22 instead of 0.31). This is just 19% of all items, while in the noise-less case 37% of items would be worth investigating. As noise becomes worse the selection for a watchlist should become stricter: a really cursory inspection should not lead to insertion unless it looks really relevant.

Here we used a mild Gaussian distribution. In term of danger, I think people or things are more likely to be lognormal distributed since it is a product of many relatively independent factors. Using lognormal x and y leads to a situation where there is a maximum utility for some threshold. This is likely a problematic model, but clearly the shape of the distributions matter a lot for where the threshold should be.

Note that having huge resources can be a bane: if you build your watchlist from the top priority down as long as you have budget or manpower, the lower priority (but still above threshold!) entries will be more likely to be a waste of time and effort. The average utility will decline.

Predictive validity matters more?

In any case, a cursory and cheap decision process is going to give so many so-so evaluations that one shouldn’t build the watchlist on it. Instead one should aim for a series of filters of increasing sophistication (and cost) to wash out the relevant items from the dross.

But even there there are pitfalls, as this paper looking at the pharma R&D industry shows:

We find that when searching for rare positives (e.g., candidates that will successfully complete clinical development), changes in the predictive validity of screening and disease models that many people working in drug discovery would regard as small and/or unknowable (i.e., an 0.1 absolute change in correlation coefficient between model output and clinical outcomes in man) can offset large (e.g., 10 fold, even 100 fold) changes in models’ brute-force efficiency.

Just like for drugs (an example where the watchlist is a set of candidate compounds), it might be more important for terrorist watchlists to aim for signs with predictive power of being a bad guy, rather than being correlated with being a bad guy. Otherwise anti-terrorism will suffer the same problem of declining productivity, despite ever more sophisticated algorithms.

Scientific progress goes zig-zag

I recently nerded out about high-energy proton interaction with matter, enjoying reading up on the Bethe equation at the Particle Data Group review and elsewhere. That got me to look around at the PDL website, which is full of awesome stuff – everything from math and physics reviews to data for the most obscure “particles” ever, plus tests of how conserved the conservation laws are.

The first thing that strikes the viewer is that they have moved a fair bit, including often being far outside the original error bars. 6 of them have escaped them. That doesn’t look very good for science!

Fortunately, it turns out that these error bars are not 95% confidence intervals (the most common form in many branches of science) but 68.3% confidence intervals (one standard deviation, if things are normal). That means having half of them out of range is entirely reasonable! On the other hand, most researchers don’t understand error bars (original paper), and we should be able to do much better.

The PDG state:

Sometimes large changes occur. These usually reflect the introduction of significant new data or the discarding of older data. Older data are discarded in favor of newer data when it is felt that the newer data have smaller systematic errors, or have more checks on systematic errors, or have made corrections unknown at the time of the older experiments, or simply have much smaller errors. Sometimes, the scale factor becomes large near the time at which a large jump takes place, reflecting the uncertainty introduced by the new and inconsistent data. By and large, however, a full scan of our history plots shows a dull progression toward greater precision at central values quite consistent with the first data points shown.

Overall, kudos to PDG for showing the history and making it clearer what is going on! But I do not agree it is a dull progression.

Zigzag to truth

The locus classicus for histories of physical constants being not quite a monotonic march towards truth is Max Henrion and Baruch Fischhoff. Assessing uncertainty in physical constants. American Journal of Physics 54, 791 (1986); doi: 10.1119/1.14447. They discuss the problem of people being overconfident and badly calibrated, and then show the zigzagging approach to current values:

Note that the shifts were far larger than the estimated error bars. The dip in the 1930s and 40s even made some physicists propose that c could be changing over time. Overall Henrion and Fischhoff find that physicists have been rather overconfident in their tight error bounds on their measurements. The approach towards current estimates is anything but dull, and hides many amusing historical anecdotes.

Stories like this might have been helpful; it is notable that the PDG histories on the right, for newer constants, seem to stay closer to the present value than the longer ones to the left. Maybe this is just because they have not had the time to veer off yet, but one can be hopeful.

Still, even if people are improving this might not mean the conclusions stay stable or approach truth monotonically. A related issue is “negative learning”, where more data and improved models make the consensus view of a topic move in the wrong direction: Oppenheimer, M., O’Neill, B. C., & Webster, M. (2008). Negative learning. Climatic Change, 89(1-2), 155-172. Here the problem is not just that people are overconfident in how certain they can be about their conclusions, but also that there is a bit of group-think, plus that the models change in structure and are affected in different ways by the same data. They point out how estimates of ozone depletion oscillated, or the consensus on the stability of climate has shifted from oscillatory (before 1968) towards instability (68-82), towards stability (82-96), and now towards instability again (96-06). These problems are not due to mere irrationality, but the fact that as we learn more and build better models these incomplete but better models may still deviate strongly from the ground truth because they miss some key component.

Noli fumare

This is related to what Nick Bostrom calls the “data fumes” problem. Early data will be fragmentary and explanations uncertain – but the data points and their patterns are very salient, just as the early models, since there is nothing else. So we begin to anchor on them. Then new data arrives and the models improve… and the old patterns are revealed as statistical noise, or bugs in the simulation or plotting routine. But since we anchored on them, we are unlikely to update as strongly towards the new most likely estimates. Worse, accommodating a new model takes mental work; our status quo bias will be pushing against the update. Even if we do accommodate the new state, things will likely change more – we may well end up either with a view anchored on early noise, or assume that the final state is far more uncertain than it actually is (since we weigh the early jumps strongly because of their saliency).

This is of course why most people prefer to believe a charismatic diet cultleader expert rather than trying to dig through 70 years of messy, conflicting dietary epidemiology.

Here is a simple example where an agent is trying to do a maximum likelihood estimation of a Gaussian distribution with mean 1 and variance 1, but is hamstrung by giving double weight to the first 9 data points:

It is not hard to complicate the model with anchoring/recency/status quo bias (estimates get biased towards previous estimates), or that early data points are more polluted by differently distributed noise. Asymmetric error checking (you will look for bugs if results deviate from expectation and hence often find such bugs, but not look for bugs making your results closer to expectation) is another obvious factor for how data fumes can get integrated in models.

The problem with data fumes is that it is not easy to tell when you have stabilized enough to start trusting the data. It is even messier when the inputs are results generated by your own models or code. I like to approach it by using multiple models to guesstimate model error: for example, one mathematical model on paper and one Monte Carlo simulation – if they don’t agree, then I should disregard either answer and keep on improving.

Even when everything seems to be fine there may be a big crucial consideration one has missed. The Turing-Good estimator gives another way of estimating the risk of that: if you have acquired $N$ data points and seen $K$ big surprises (remember that the first data point counts as one!), then the probability of a new surprise for your next data point is $\approx K/N$. So if you expect $M$ data points in total, when $K(M-N)/N \ll 1$ you can start to trust the estimates… assuming surprises are uncorrelated etc. Which you will not be certain about. The progression towards greater precision may be anything but dull.