Monday, March 30, 2020

Robin Hanson: Variolation May Cut Covid19 Deaths 3-30X

Variolation May Cut Covid19 Deaths 3-30X. Robin Hanson. Overcoming Bias, March 30, 2020.

(Here I try to put my recent arguments together into an integrated essay, suitable for recommending to others.)

When facing a new pandemic, the biggest win is to end it fast, so that few ever suffer. This prize makes it well worth trying hard to trace, test, and isolate those near the first few cases. Alas, for Covid-19 and the world, this has mostly failed, though not yet everywhere.

The next biggest win is to find a cheap effective treatment, such as a vaccine. And while hope remains for an early win, this looks to be years away. To keep most from getting infected, at this point the West must apparently develop and long maintain unprecedented expansions in border controls, testing, tracing, and privacy invasions, and perhaps also non-home isolation of suspected cases. Alas, these ambitious plans must be implemented by the same governments that have so far failed us badly.

Yes, there remains hope here, which should be pursued. But we also need a Plan B; what if most will eventually be infected without a treatment? The usual answer is “flatten the curve,” via more social distance to lower the average of (and increase the variance of) infection rates, so that more can access limited medical resources. Such as ventilators, which cut deaths by <¼, since >¾ of patients on them die.

However, extreme “lockdowns”, which isolate most everyone at home, not only limit freedoms and strangle the economy, they also greatly increase death rates. This is because infections at home via close contacts tend to come with higher initial virus doses, in contrast to the smaller doses you might get from, say, a public door handle. As soon as your body notices an infection, it immediately tries to grow a response, while the virus tries to grow itself. From then on, it is a race to see which can grow biggest fastest. And the virus gets a big advantage in this race if its initial dose of infecting virus is larger.

This isn’t just a theory. The medical literature consistently finds strong relations, in both animals and humans, between initial virus dose and symptom severity, including death. The most directly relevant data is on SARS and measles, where natural differences in doses were associated with factors of 3 and 14 in death rates, and in smallpox, where in the 1700s low “variolation” doses given on purpose cut death rates by a factor of 10 to 30. For example, variolation saved George Washington’s troops at Valley Forge.

Early on, it can be worth paying such high costs to end a pandemic. But once a pandemic seems likely to eventually infect most everyone, it becomes less clear whether lockdowns are a net win. However, the dose effect that lockdowns exacerbate, by increasing dose size, also offers a huge opportunity to slash deaths, via voluntary infection with very low doses.

Just as replacing accidental smallpox infections with deliberate low dose infections cut smallpox deaths by a factor of 10 to 30, a factor of 3-30 is plausible for Covid19 death rate cuts due to replacing accidental Covid19 infections with deliberate small dose infections. Observed mortality differences due to natural dose variations give only a lower bound on what is feasible via controlled doses. Of course we can’t be sure until we get more direct evidence. But systematic variolation experiments involving at most a few thousand volunteers seem sufficient to get evidence not only on death rates, but also on ideal infection doses and methods, and on the value of complementary drugs that slow viral replication (e.g., remdesivir).

This dose size advantage adds to several other substantial advantages of variolation. Not only does it offer controlled conditions for studying disease progression, and for training medical personnel, it can also help ensure consistent staffing of critical workers, by spacing out their infections.

Furthermore, the combination of variolation with immediate isolation until recovery “flattens the curve,” by spreading out medical demand over time, and also adding to the herd immunity that usually ends a pandemic. So even without a death rate cut due to lower doses, this strategy produces a net social gain.

This last claim may sound counter-intuitive, but it has in fact recently been confirmed in three independently developed simulations. For example, in a simulation where old and sick people are selected for isolation, while only the young and healthy are eligible for variolation, there are 40% fewer life years lost, compared to no variolation and random selection for isolation. Each variolation volunteer suffers only an additional 0.20% chance of death to save a random other person from a 6.5% chance. And these simulations ignore any benefits of low doses; they hold constant the infection and death rates, and the total quantity of social isolation, and thus expense.

Of course, if low doses cut death rates by a factor of two or more, variolation volunteers would actually cut their chance of death, perhaps greatly. Yes, the first few thousand volunteers could be less sure of such gains, but they could be compensated for this risk, just as we now consider compensating subjects in vaccine trials using live Covid19 viruses. We could pay variolation volunteers cash, offer their loved ones priority medical care, certify them as safe for work and social gatherings, and honor them like soldiers selected for their elite features who take risks to produce community gains.

So the scenario is this: Variolation Villages welcome qualified volunteers. Friends and family can enter together, and remain together. A cohort enters together, and is briefly isolated individually for as long as it takes to verify that they’ve been infected with a very small dose of the virus. They can then interact freely with each other, but not leave the village until tests show they have recovered.

In Variolation Village, volunteers have a room, food, internet connection, and full medical care. Depending on available funding from government or philanthropic sources, volunteers might either pay to enter, get everything for free, or be paid a bonus to enter. Health plans of volunteers may even contribute to the expense.

Those who work in medicine or critical infrastructure seem especially valuable candidates for early variolation; volunteers might be offered larger bonuses. Once they have recovered, they are more surely available to work near the pandemic peak, and can more easily risk social contact at work.

Note that this strategy of variolation plus isolation requires no government support, nor loss of personal freedom, just the sort of legal permission sometimes given to administrators and volunteers of vaccine trials. And this comparison with vaccine trial policy can be emphasized to those tempted to see this policy as repulsive. Variolation policy offers similar social gains, and may require similar voluntary personal sacrifices.

Note also that there is no minimum scale required to make this policy beneficial. Even variolation of only a few is still a social gain compared to none at all. A small early trial could generate much useful attention and discussion regarding this strategy, to inspire application in this and future pandemics. Furthermore, the optimal time to stop this practice for personal reasons is probably close to the optimal time to stop for social reasons, so choice of stopping date needn’t be heavily regulated.

Some fear that it is now too late to consider variolation, as the pandemic peak may be only a few weeks away. But lockdowns may succeed in substantially slowing Covid19 growth, and we may then be in for many months or years of alternating local waves of suppression and reappearance. Furthermore, if low doses cut death rates enough, variolation can make sense even at the pandemic peak, when medical resources are stretched most thin. For example, for a factor of 3 cut in death rates, variolation replaces three sick patients with one similarly sick patient, lowering total medical demand.

As variolation doesn’t much change the total number who are ever infected, it doesn’t give the virus more total chances to evolve. In fact, while accidental infections risk selection for versions that infect people more easily, voluntary infections avoid this problematic effect.

While you might think policy wonks would be eager to cut Covid19 death rates by a factor of 3-30, few have so far been attracted to discuss or pursue this concept. It seems to push the wrong buttons in many people. So if you are a rare exception who finds the concept plausible, you can get a disproportionate policy leverage by working on a neglected important option. You might help in one of these areas:

[more text at the link above]

We have much work to do if this Plan B is to be ready when needed.

Adults severely underestimate their absolute and relative fatality risk if infected with SARS-CoV

Niepel, Christoph, Dirk Kranz, Francesca Borgonovi, and Samuel Greiff. 2020. “Sars-cov-2 Fatality Risk Perception in US Adult Residents.” PsyArXiv. March 30. doi:10.31234/

Abstract: Our study presents time-critical empirical results on the SARS-CoV-2 fatality risk perception of 1182 US adult residents stratified for age and gender. Given the current epidemiological figures, our findings suggest that many US adult residents severely underestimate their absolute and relative fatality risk if infected with SARS-CoV-2. These results are worrying because risk perception, as our study suggests, relates to self-reported actual or intended behavior that can reduce SARS-CoV-2 transmission rates.

Animals benefit from numerical competence (foraging, navigating, hunting, predation avoidance, social interactions, & reproductive activities); internal number representations determine how they perceive stimulus magnitude

The Adaptive Value of Numerical Competence. Andreas Nieder. Trends in Ecology & Evolution, March 30 2020.

*  Numerical competence, the ability to estimate and process the number of objects and events, is of adaptive value.
*  It enhances an animal’s ability to survive by exploiting food sources, hunting prey, avoiding predation, navigating, and persisting in social interactions. It also plays a major role in successful reproduction, from monopolizing receptive mates to increasing the chances of fertilizing an egg and promoting the survival chances of offspring.
*  In these ecologically relevant scenarios, animals exhibit a specific way of internally representing numbers that follows the Weber-Fechner law.
*  A framework is provided for more dedicated and quantitative analyses of the adaptive value of numerical competence.

Abstract: Evolution selects for traits that are of adaptive value and increase the fitness of an individual or population. Numerical competence, the ability to estimate and process the number of objects and events, is a cognitive capacity that also influences an individual’s survival and reproduction success. Numerical assessments are ubiquitous in a broad range of ecological contexts. Animals benefit from numerical competence during foraging, navigating, hunting, predation avoidance, social interactions, and reproductive activities. The internal number representations determine how animals perceive stimulus magnitude, which, in turn, constrains an animal’s spontaneous decisions. These findings are placed in a framework to provide for a more quantitative analysis of the adaptive value and selection pressures of numerical competence.

Keywords: quantitynumberWeber-Fechner lawproportional processingultimate causesanimal cognition

UK: Inequality in socio-emotional skills has increased across cohorts, especially for boys and at the bottom of the distribution

Inequality in socio-emotional skills: A cross-cohort comparison. Orazio Attanasio et al. Journal of Public Economics, March 30 2020, 104171.

Abstract: We examine changes in inequality in socio-emotional skills very early in life in two British cohorts born 30 years apart. We construct comparable scales using two validated instruments for the measurement of child behaviour and identify two dimensions of socio-emotional skills: ‘internalising’ and ‘externalising’. Using recent methodological advances in factor analysis, we establish comparability in the inequality of these early skills across cohorts, but not in their average level. We document for the first time that inequality in socio-emotional skills has increased across cohorts, especially for boys and at the bottom of the distribution. We also formally decompose the sources of the increase in inequality and find that compositional changes explain half of the rise in inequality in externalising skills. On the other hand, the increase in inequality in internalising skills seems entirely driven by changes in returns to background characteristics. Lastly, we document that socio-emotional skills measured at an earlier age than in most of the existing literature are significant predictors of health and health behaviours. Our results show the importance of formally testing comparability of measurements to study skills differences across groups, and in general point to the role of inequalities in the early years for the accumulation of health and human capital across the life course.

JEL classification: J13J24I14I24C38
Keywords: InequalitySocio-emotional skillsCohort studiesMeasurement invariance

Our results imply that the ability to utilize the enhanced information of a face to recognize familiar faces may develop aged around 7 months of age

Infants’ recognition of their mothers’ faces in facial drawings. Megumi Kobayashi  Ryusuke Kakigi  So Kanazawa  Masami K. Yamaguchi. Developmental Psychobiology, March 29 2020.

Abstract: This study examined the development of ability to recognize familiar face in drawings in infants aged 6–8 months. In Experiment 1, we investigated infants’ recognition of their mothers’ faces by testing their visual preference for their mother’s face over a stranger’s face under three conditions: photographs, cartoons produced by online software that simplifies and enhances the contours of facial features of line drawings, and veridical line drawings. We found that 7‐ and 8‐month‐old infants showed a significant preference for their mother’s face in photographs and cartoons, but not in veridical line drawings. In contrast, 6‐month‐old infants preferred their mother’s face only in photographs. In Experiment 2, we investigated a visual preference for an upright face over an inverted face for cartoons and veridical line drawings in 6‐ to 8‐month‐old infants, finding that infants aged older than 6 months showed the inversion effect in face preference in both cartoons and veridical line drawings. Our results imply that the ability to utilize the enhanced information of a face to recognize familiar faces may develop aged around 7 months of age.