An algorithm that can spot cause and effect could supercharge medical AI

How do we make better use of this piecemeal information? Computers are great at spotting patterns—but that’s just correlation. In the last few years, computer scientists have invented a handful of algorithms that can identify causal relations within single data sets. But focusing on single data sets is like looking through keyholes. What’s needed is a way to take in the whole view.

Our world is awash in bullshit health claims. These scientists want to train kids to spot them.

Screen Shot 2016-09-28 at 10.07.44 AM.png

“We are trying to teach children that stories are usually an unreliable basis for assessing the effect of treatments,” Nsangi explained, adding that stories amount to anecdotal evidence. The kids are also learning to watch out for the perverting effects of conflicts of interest, and to recognize that all treatments carry both harms and benefits and that large, dramatic effects from a treatment are really, really rare.

Sorry, There’s Nothing Magical About Breakfast

Good reviews of all the observational research note the methodological flaws in this domain, as well as the problems of combining the results of publication-bias-influenced studies into a meta-analysis. The associations should be viewed with skepticism and confirmed with prospective trials.

“Few randomized controlled trials exist. Those that do, although methodologically weak like most nutrition studies, don’t support the necessity of breakfast.”

This article raises some interesting points about what is “good science” and also raises the issue of how hard it is to “know” things when it comes to nutrition. Though not the main focus, the article also touches upon whether the funding source of a study affects the conclusion and how ethical it is to fund a self serving study.

Researchers find greater incidence of cancer in tall people

Here is an interesting story about research that found a correlation between height and incidence of cancer. Generally I despise health news covered by television news because they generally butcher the science or oversell the conclusions that are appropriate based on the research. What I like about this story is that if you listen carefully there are elements here relevant to our class.

  • This study is valid because it followed 5.5 million people over a long period of time. Large sample size and longer time give the study more weight.
  • The experts say that there isn’t an understood cause just yet simply a correlation between height and incidence of cancer.
  • They go on to speculate about how height might cause increased risk but they acknowledge that they don’t know. Is it because you have more cells in your body that can go bad? Is it because you have more skin exposed to the sun? Growth hormone? Maybe. They don’t know that any of those theories are true but they seem to have found a correlation in a study in one country. When doing scientific research, especially about health, skepticism is important.

One last thing to think about is whether a study done in Sweden can be applicable to people in general. Is there something unique about the environment there? Diet? Genetics? Would these results still be true if done in a country that gets more sunlight year round?

Considering all these factors should illustrate to you why it’s so hard to come to conclusions about human health.