The Adaptable Animal: The Evolutionary Context and Practical Tactics for Building Your Resilience

"The government should've brought in [universities and researchers] from the outset" @GregClarkMP, chair of the @CommonsSTC, talks to #5LiveBreakfast about some of the lessons to learn about how the UK handled #COVID19.

10.1136/bmj.m1992

Just 0.27% of people—one in 370—are estimated to have had covid-19 outside of hospitals and care homes in England in the past two weeks, say the preliminary results of a snapshot survey published by the Office for National Statistics (ONS).1The figures were described by England’s deputy chief medical officer, Jonathan Van-Tam, as “quite a low level of infection in the community,” at the government’s daily briefing on 14 May.The survey results come as studies from Spain and France indicate that just 5.0% and 4.4% of their populations respectively have ever contracted covid-19, indicating that most people may still be susceptible to infection.

Covid-19: Surveys indicate low infection level in community

These maps come from the brilliant Wellcome Collection: https://wellcomecollection.org/works/ And I wrote a similar threat with full sources back in 2018 http://spatial.ly/2019/03/mapping-and-visualising-cholera-data/… /ends.

...and also looking to the geology of London in later years.

Of course, they were an essential rebuke to the miasma theory and enabled the attention to turn to London's water supply. It turned out that certain suppliers were responsible for higher deaths than others...

So Snow had his work cut out...he pounded the streets getting detailed on the ground data - not just headline figures for cities. The black bars on his map show a death from cholera. I think it has a humanity that Farr's visualisations lack...

And others made scary maps showing the "cholera mist" as it spread across London...

Farr's charts make a particularly beautiful case for the miasma theory...

People wanted answers so William Farr stepped up. He was a rigorous statistician & excellent data visualiser. His (and the prevailing) theory was that cholera was spread by bad smells "miasma". London was a smelly place and outbreaks peaked with hot weather when it was smelliest.

Still no sign of Snow's map at this point...but it was getting serious in London...

As it arrived in Britain towns and cities were hardest hit...here's the extent of the epidemic in 1849.

Of course, cholera remains a global issue killing tens of thousands a year. Here's an early map showing its spread from India to Britain...we've seen so many recent versions of this with COVID-19.

John Snow's map of cholera looked as dull as (cholera filled) dishwater compared to his competitors... His brilliance was a solid data collection & then a simple map presenting what he knew. Each death marked in black and white. Here's a lesson for COVID-19 dataviz...

10.1103/PhysRevE.101.052310

The average fraction of infected nodes, in short the prevalence, of the Markovian ɛ<math xmlns="http://www.w3.org/1998/Math/MathML"><mi>ɛ</mi></math>-SIS (susceptible-infected-susceptible) process with small self-infection rate ɛ>0<math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>ɛ</mi><mo>&gt;</mo><mn>0</mn></mrow></math> exhibits, as a function of time, a typical “two-plateau” behavior, which was first discovered in the complete graph KN<math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>K</mi><mi>N</mi></msub></math>. Although the complete graph is often dismissed as an unacceptably simplistic approximation, its analytic tractability allows to unravel deeper details, that are surprisingly also observed in other graphs as demonstrated by simulations. The time-dependent mean-field approximation for KN<math xmlns="http://www.w3.org/1998/Math/MathML"><msub><mi>K</mi><mi>N</mi></msub></math> performs only reasonably well for relatively large self-infection rates, but completely fails to mimic the typical Markovian ɛ<math xmlns="http://www.w3.org/1998/Math/MathML"><mi>ɛ</mi></math>-SIS process with small self-infection rates. While self-infections, particularly when their rate is small, are usually ignored, the interplay of nodal self-infection and spread over links may explain why absorbing processes are hardly observed in reality, even over long time intervals.

Time dependence of susceptible-infected-susceptible epidemics on networks with nodal self-infections

10.1103/PhysRevE.101.052211

In this paper, we broaden the master stability function approach to study the stability of the synchronization manifold in complex networks of stochastic dynamical systems. We provide necessary and sufficient conditions for exponential stability that allow us to discriminate the impact of noise. We observe that noise can be beneficial for synchronization when it diffuses evenly in the network. On the contrary, an excessively large amount of noise only acting on a subset of the node state variables might have disruptive effects on the network synchronizability. To demonstrate our findings, we complement our theoretical derivations with extensive simulations on paradigmatic examples of networks of noisy systems.

Stochastic master stability function for noisy complex networks

2005.08050

We consider information diffusion on Web-like networks and how random walks can simulate it. A well-studied problem in this domain is Partial Cover Time, i.e., the calculation of the expected number of steps a random walker needs to visit a given fraction of the nodes of the network. We notice that some of the fastest solutions in fact require that nodes have perfect knowledge of the degree distribution of their neighbors, which in many practical cases is not obtainable, e.g., for privacy reasons. We thus introduce a version of the Cover problem that considers such limitations: Partial Cover Time with Budget. The budget is a limit on the number of neighbors that can be inspected for their degree; we have adapted optimal random walks strategies from the literature to operate under such budget. Our solution is called Min-degree (MD) and, essentially, it biases random walkers towards visiting peripheral areas of the network first. Extensive benchmarking on six real datasets proves that the---perhaps counter-intuitive strategy---MD strategy is in fact highly competitive wrt. state-of-the-art algorithms for cover.

Exploring Low-degree Nodes First Accelerates Network Exploration

2005.08074

Universal spectral properties of multiplex networks allow us to assess the nature of the transition between disease-free and endemic phases in the SIS epidemic spreading model. In a multiplex network, depending on a coupling parameter, pp, the inverse participation ratio (IPRIPR) of the leading eigenvector of the adjacency matrix can be in two different structural regimes: (i) layer-localized and (ii) delocalized. Here we formalize the structural transition point, p∗p^*, between these two regimes, showing that there are universal properties regarding both the layer size nn and the layer configurations. Namely, we show that IPR∼n−δIPR \sim n^{-\delta}, with δ≈1\delta\approx 1, and revealed an approximately linear relationship between p∗p^* and the difference between the layers' average degrees. Furthermore, we showed that this multiplex structural transition is intrinsically connected with the nature of the SIS phase transition, allowing us to both understand and quantify the phenomenon. As these results are related to the universal properties of the leading eigenvector, we expect that our findings might be relevant to other dynamical processes in complex networks.

Universality of eigenvector delocalization and the nature of the SIS phase transition in multiplex networks

Invest in Open Infrastructure (IOI) was launched to create a strategic, global body dedicated to furthering a network of open, interoperable community-led and -supported infrastructure to advance scholarship, research, and education.  We are currently ramping up efforts to help support university decision makers, consortia, and funders globally to sustain research and knowledge sharing amidst these uncertain times. 

Open Infrastructure in times of crisis: How IOI can help

In response to the COVID-19 outbreak, INGSA has created this information hub to aggregate and share the resources and discussions relating to how science advice and evidence functions in emergencies. We are looking for experts and practitioners to write commentary and analysis, share resources and opportunities, and provide input into our national policy-making tracker.

Science Advice and COVID-19

10.31234/osf.io/xn7sr

We present an online platform, called BeeMe, designed to test the current boundaries of Internet collective action and problem solving. BeeMe allows a scalable internet crowd of online users to collectively control the actions of a human surrogate acting in physical space. BeeMe demonstrates how intelligent goal-oriented decision-making can emerge from large crowds in quasi real-time. We analyzed data collected from a global BeeMe live performance that involved thousands of individuals, collectively solving a sci-fi Internet mystery. We study simple heuristic algorithms that read in users' chat messages and output human actionable commands representing majority preferences, and compare their performance to the behavior of a human operator solving the same task. Results show that simple heuristics can achieve near-human performance in interpreting the democratic consensus. When human and machine's output differ, the discrepancy is often due to human bias favoring non-representative views. We discuss our results in light of previous work and the contemporary debate on democratic digital systems.

Real-time Internet Control of Situated Human Agents

10.1101/2020.05.13.092619

During the COVID-19 pandemic, SARS-CoV-2 infected millions of people and claimed hundreds of thousands of lives. Virus entry into cells depends on the receptor binding domain (RBD) of the SARS-CoV-2 spike protein (S). Although there is no vaccine, it is likely that antibodies will be essential for protection. However, little is known about the human antibody response to SARS-CoV-21-5. Here we report on 68 COVID-19 convalescent individuals. Plasmas collected an average of 30 days after the onset of symptoms had variable half-maximal neutralizing titers ranging from undetectable in 18% to below 1:1000 in 78%, while only 3% showed titers >1:5000. Antibody cloning revealed expanded clones of RBD-specific memory B cells expressing closely related antibodies in different individuals. Despite low plasma titers, antibodies to distinct epitopes on RBD neutralized at half-maximal inhibitory concentrations (IC50s) as low as few ng/mL. Thus, most convalescent plasmas obtained from individuals who recover from COVID-19 without hospitalization do not contain high levels of neutralizing activity. Nevertheless, rare but recurring RBD-specific antibodies with potent antiviral activity were found in all individuals tested, suggesting that a vaccine designed to elicit such antibodies could be broadly effective.

Convergent Antibody Responses to SARS-CoV-2 Infection in Convalescent Individuals

10.13140/RG.2.2.27042.66243

Many digital solutions mainly involving Bluetooth technology are being proposed for Contact Tracing Apps (CTA) to reduce the spread of COVID-19. Concerns have been raised regarding privacy, consent, uptake required in a given population, and the degree to which use of CTAs can impact individual behaviours. The introduction of a new CTA alone will not contain COVID-19. The best-case scenario for uptake requires between 90 and 95% of the entire population for containment. This does not factor in any loss due to people dropping out or device incompatibility or that only 79% of the population own a smartphone, with less than 40% in the over-65 age group. Hence, the best-case scenario is beyond that which could conceivably be achieved. We propose to build on some of the digital solutions already under development, with the addition of a Bayesian network model that predicts likelihood for infection supplemented by traditional symptom and contact tracing. When combined with freely available COVID-19 testing with results in 24 hours or less, an effective communication strategy and social distancing, this solution can have a very beneficial effect on containing the spread of this pandemic.

The fundamental limitations of COVID-19 contact tracing methods and how to resolve them with a Bayesian network approach

The Supplementary Figures can be found here as a public link.

Whether this immunity is relevant in influencing clinical outcomes is unknown, but it is tempting to speculate that the crossreactive CD4+ T cells may be of value in protective immunity, based on SARS and flu data.

Crossreactive T cells are also relevant for vaccine development, as cross-reactive immunity could influence responsiveness to candidate vaccines

Detecting SARS2-reactive T cells in ~50% of unexposed people suggests cross-reactive T cell recognition between circulating ‘common cold’ coronaviruses and SARS-CoV-2. This might influence susceptibility to COVID-19 disease.

Additionally, any potential for crossreactive immunity from other coronaviruses has been predicted by epidemiologists to have significant implications for the pandemic going forwards. We detected SARS-CoV-2-reactive CD4+ T cells in ~50% of unexposed individuals.

We specifically chose to study people who had an average COVID19 disease course—non-hospitalized—to provide a solid benchmark for what a normal immune response to SARS2 looks like.

CD4+ T cell responses to spike, the main target of most vaccine efforts, were robust and correlated with the magnitude of the anti-SARS-CoV-2 IgG and IgA titers. Again, good news.

In our study, 100% of COVID-19 cases made antibodies. 100% of COVID-19 cases made CD4 T cells. 70% of COVID-19 cases made measurable CD8 T cells. We believe these findings are good news, and consistent with normal antiviral immunity.

There has been a huge amount of uncertainty about immunity to SARS2—both in the context of COVID19 disease pathogenesis and in the context of how to develop a good vaccine.

This is good news in multiple ways, for coronavirus vaccine development, understanding disease, and even modeling the future course of the pandemic.

Our new paper rapidly studied T cell + antibody immune responses in average COVID-19 cases. This is good news in multiple ways, including coronavirus vaccines.

10.1098/rspa.2019.0772

Network topologies can be highly non-trivial, due to the complex underlying behaviours that form them. While past research has shown that some processes on networks may be characterized by local statistics describing nodes and their neighbours, such as degree assortativity, these quantities fail to capture important sources of variation in network structure. We define a property called transsortativity that describes correlations among a node’s neighbours. Transsortativity can be systematically varied, independently of the network’s degree distribution and assortativity. Moreover, it can significantly impact the spread of contagions as well as the perceptions of neighbours, known as the majority illusion. Our work improves our ability to create and analyse more realistic models of complex networks.

The transsortative structure of networks

2005.07427

Detecting anomalies in dynamic graphs is a vital task, with numerous practical applications in areas such as security, finance, and social media. Previous network embedding based methods have been mostly focusing on learning good node representations, whereas largely ignoring the subgraph structural changes related to the target nodes in dynamic graphs. In this paper, we propose StrGNN, an end-to-end structural temporal Graph Neural Network model for detecting anomalous edges in dynamic graphs. In particular, we first extract the hh-hop enclosing subgraph centered on the target edge and propose the node labeling function to identify the role of each node in the subgraph. Then, we leverage graph convolution operation and Sortpooling layer to extract the fixed-size feature from each snapshot/timestamp. Based on the extracted features, we utilize Gated recurrent units (GRUs) to capture the temporal information for anomaly detection. Extensive experiments on six benchmark datasets and a real enterprise security system demonstrate the effectiveness of StrGNN.

Structural Temporal Graph Neural Networks for Anomaly Detection in Dynamic Graphs

10.1136/bmj.m1932

The UK government and its advisers were confident that theywere “well prepared” when covid-19 swept East Asia. Thefour-pronged plan of 3 March to contain, delay, research, andmitigate was supported by all UK countries and backed, theyclaimed, by science.1 With over 30 000 hospital and communitydeaths by 12 May, where did the plan go wrong?2 What was therole of public health in the biggest public health crisis since theSpanish flu of 1918? And what now needs to be done?What is clear is that the UK’s response so far has neither beenwell prepared nor remotely adequate (see infographic). Theweakness of the preparations was exposed in 2016 by ExerciseCygnus, a pandemic simulation, and the necessary remedialsteps were not taken.3 On 30 January, the World HealthOrganization declared a public health emergency of internationalconcern and governments were urged to prepare for globalspread of covid-19 from East Asia.4 Detailed case studiesfollowed showing the need for high levels of mechanicalventilation and high death rates.5 6 But the UK ignored thesewarnings

The UK’s public health response to covid-19

For the sake of both science and action in the COVID-19 pandemic, we need collaboration among specialists, not sects.

Good Science Is Good Science

10.1136/bmj.m1847

The medical research world is responding to the covid-19 pandemic at breathtaking speed. There has been a maelstrom of global research, with mixed consequences. Positives include the greater provision of open access to covid-19 studies, some increased collaboration, expedited governance and ethics approvals of new clinical studies, and wider use of preprints. But many problems have become evident. Before the pandemic, it was estimated that up to 85% of research was wasted because of poor questions, poor study design, inefficiency of regulation and conduct, and non or poor reporting of results.1 Many of these problems are amplified in covid-19 research, with time pressures and inadequate research infrastructure contributing.

Waste in covid-19 research

Download the most recent estimates here

COVID-19 estimate downloads

The Swedish Intensive Care Register receives data reported on the cases of Covid-19 that end up in Swedish intensive care units. How quickly SIR receives data depends on local reporting procedures and local IT systems. The age range between the youngest and the oldest admitted to the intensive care unit due to Covid-19 is large. At present, the majority are men.

COVID-19 in Swedish intensive care

PS. People interested in learning more about overconfidence among scientist-experts might want to check this piece out:

Anyway, excellent journalism by @MariaGuntherA and @MarrisW of @dagensnyheter / @dn_grafik /fin

Scientist overconfidence is a massive problem. In the short run, it undercuts efforts to use science to inform policy; in the long run, it reduces trust in science in general. We can and should do better /10

The people involved in these forecasts expressed themselves with *way* more confidence than what was justified at the time. This was an unforced error on their behalf /9

We should be less tolerant of overconfidence in particular and a lack of epistemic humility in general. A true expert would have known ahead of time just how much uncertainty was involved in their forecasts and expressed themselves accordingly /8

We should be tolerant of mistaken projections. These are incredibly difficult prediction tasks. The modellers here were trying to be useful, and they were working under great time pressure /7

From a sociology of science perspective, we should expect few modellers to admit having made mistakes: based on @PTetlock's research we should expect claims to the effect that they were "almost right." So far I haven't seen one saying "we were wrong." (But I could be wrong!) /6

In addition, some of these models apparently contain over 100 parameters, and would be difficult to calibrate under any conditions /5

From a philosophy of science perspective, this should not be surprising. Models work well when the underlying data-generating process is known and stable and when there has been ample time to calibrate the model. These conditions do not obtain here. /4

Around the same time, if I read their data file correctly, the IHME projected a demand of 4400, with a 95% uncertainty interval of 1400–11000. The real number is therefore way outside the interval /3

tl;dr Model-based projections drastically exaggerated the actual demand – sometimes by more than an order of magnitude. Today the number of patients in intensive care is about 450; it never exceeded 600 /2

Terrific assessment of projections of demand for Swedish ICU beds. The first two panels are model-based projections by academics; the third is a simple extrapolation by the public-health authority; the fourth is the actual outcome /1

10.1037/dec0000111

Two consistent findings from the study of the fit between judgment of performance and actual performance are general overconfidence and the hard–easy effect, with overconfidence being higher with more difficult stimuli. These findings are based on aggregated analyses of confidence and accuracy, despite the fact that confidence judgments are individual and are provided at the item level. Furthermore, an important characteristic of item performance judgments that is ignored by traditional analyses is that the objective difficulty of any item can be estimated before it is administered to a person. We argue that traditional analyses confound possible bias in subjective estimates of the difficulty of items (i.e., confidence judgments) with variations in objective difficulty of items. We propose a multilevel approach to the analysis of confidence judgments, whereby the probability of a correct response is modeled as a function of both objective difficulty and subjectively judged difficulty. In this model, the intercept represents the possible overall bias (over- or underconfidence) in subjective difficulty judgments, after controlling for objective difficulty as well as variations across persons and items. In effect we are proposing a new, more nuanced, standard for defining calibration and identifying distinct patterns of miscalibration. We demonstrate the confounding effects of conventional aggregated analysis through synthetic examples and apply the proposed approach to the analysis of empirical data. Conventional analyses replicated the overall overconfidence and the hard–easy effect, but the item response modeling results failed to identify an overall bias in confidence judgments or a test difficulty effect.

An item response approach to calibration of confidence judgments.

10.1037/0278-7393.30.6.1167

Format dependence implies that assessment of the same subjective probability distribution produces different conclusions about over- or underconfidence depending on the assessment format. In 2 experiments, the authors demonstrate that the overconfidence bias that occurs when participants produce intervals for an uncertain quantity is almost abolished when they evaluate the probability that the same intervals include the quantity. The authors successfully apply a method for adaptive adjustment of probability intervals as a debiasing tool and discuss a tentative explanation in terms of a naive sampling model. According to this view, people report their experiences accurately, but they are naive in that they treat both sample proportion and sample dispersion as unbiased estimators, yielding small bias in probability evaluation but strong bias in interval production.

Subjective Probability Intervals: How to Reduce Overconfidence by Interval Evaluation.

10.1037/0033-295X.115.2.502

The authors present a reconciliation of 3 distinct ways in which the research literature has defined overconfidence: (a) overestimation of one's actual performance, (b) overplacement of one's performance relative to others, and (c) excessive precision in one's beliefs. Experimental evidence shows that reversals of the first 2 (apparent underconfidence), when they occur, tend to be on different types of tasks. On difficult tasks, people overestimate their actual performances but also mistakenly believe that they are worse than others; on easy tasks, people underestimate their actual performances but mistakenly believe they are better than others. The authors offer a straightforward theory that can explain these inconsistencies. Overprecision appears to be more persistent than either of the other 2 types of overconfidence, but its presence reduces the magnitude of both overestimation and overplacement.

The trouble with overconfidence.

10.1037/0033-295X.101.3.519

Two empirical judgment phenomena appear to contradict each other. In the revision-of-opinion literature, subjective probability (SP) judgments have been analyzed as a function of objective probability (OP) and generally have been found to be conservative, that is, to represent underconfidence. In the calibration literature, analyses of OP (operationalized as relative frequency correct) as a function of SP have led to the opposite conclusion, that judgment is generally overconfident. Reanalysis of 3 studies shows that both results can be obtained from the same set of data, depending on the method of analysis. The simultaneous effects are then generated and factors influencing them are explored by means of a model that instantiates a very general theory of how SP estimates arise from true judgments perturbed by random error. Theoretical and practical implications of the work are discussed.

Simultaneous over- and underconfidence: The role of error in judgment processes.

We can emerge from this crisis a better world, if we act quickly and jointly, writes Professor Klaus Schwab. The changes we have already seen in response to COVID-19 prove that a reset of our economic and social foundations is possible. This is our best chance to instigate stakeholder capitalism - and here's how it can be achieved.

Now is the time for a 'great reset'

The rich have a more optimistic financial outlook than those who manage their money

Wealthy investors ride the storm

10.1001/jamanetworkopen.2018.3731

Question  If commercial insurers retrospectively deny coverage for emergency department (ED) visits based on diagnoses determined to be nonemergent, what visits will be denied coverage?Findings  This cross-sectional study found that 1 insurer’s list of nonemergent diagnoses would classify 15.7% of commercially insured adult ED visits for possible coverage denial. However, these visits shared the same presenting symptoms as 87.9% of ED visits, of which 65.1% received emergency-level services.Meaning  A retrospective diagnosis-based policy is not associated with accurate identification of unnecessary ED visits and could put many commercially insured patients at risk of coverage denial.

Analysis of a Commercial Insurance Policy to Deny Coverage for Emergency Department Visits With Nonemergent Diagnoses

The pandemic shows how hard it will be to decarbonise—and creates an opportunity

Countries should seize the moment to flatten the climate curve