South Korean dataset and social contours

South Korea is publishing fairly openly a list of individuals (without until recently much concern for privacy except for witholding names) who have been infected, and the tracing the government has managed to do of the infection (this site is in Korean; hint: on Chrome right-clicking offers the option of translating the page).

Here is an example report, run through Google Translate

This site is also discussed in this talk (starting at 17:35):

It would be interesting to do some analysis on this dataset (NLP if the dataset is big), particularly the contact reports for transmission events (rather than the patient reports), in order to:

  1. assess which communities get mentioned in those reports (ensuring completeness of the thinking of the sociologists here)
  2. model how frequently the transmission hops from one community to the next, vs within each community

Both of those could be done with the ability to ignore the superspreader event of Patient 31 in South Korea (going to church, which were mapped to the heat maps available in South Korea as a suboptimal retelling of the impact of that one community).

The interest in this dataset is motivated for instance by a scientific understanding of the (limited utility) of R0 vs secondary transmissions, and a scientific understanding of the social response necessary to account for epidemics while being aware of risks such as discrimination etc (detailed here).

A social contour is a set of activities or social interactions you can undertake without the risk of causing a new outbreak.

The idea is to return to a normal life as much as possible while keeping the virus in check. So we try to develop a methodology to find social interactions we can safely undertake.

It may be safe to let the children return to school while big festivities may still be unsafe.
It may be possible to limit an outbreak to a certain community and protect an other (vulnerable) community by limiting certain interactions between communities, e.g. by creating a ‘protected hour’ at a supermarket.
It may possible to identify behaviours that carry a high risk and behaviours that carry a low risk. High risk behaviours may ask for more precautions.

It is a game of maximizing social interaction within the virological boundaries, so to find good social contours you have to take both into account. this is crucial to take into account
“Acceptabilité d’une application téléphone pour tracer les contacts
porteurs du Covid-19”
For translating deepl. I can help too

There is an English version

yes, i saw that, Tim has the different resources you gave me already

The French version is for French respondents, the German version for the German respondents, the English version for UK , etc …


in relation to people that can’t share data and tat are vulnerable we have this association i will be in touch with: