While most of us celebrated the new year, many regions in Japan were hit by an earthquake. Houses turned into rubble, and there were casualties as well. My heart and prayers go to those impacted, and I wish them a speedy recovery and relief. This morning, I came across an article (and an associated video) highlighting that birds in impacted areas behaved ” weird ” ways hours before the quake.

As a child also, I once heard from my grandmother that cattle started behaving weirdly hours before a major earthquake hit her village when she was a child. Since I had nothing more important to do, I started deep-diving into the validity of these claims. There is some relation, but not much has been done to explore that relationship. In my mind, exploring that can help us predict a natural disaster that has often claimed thousands of lives in moments. But not by studying animal behavior directly.

I came across an article on the Geological Survey website that does acknowledge that there is a relationship. Still, it shies away from putting effort into exploring that by saying: “However, much research still needs to be done on this subject. “I’m not sure if the authors are USGS employees, but shouldn’t USGS be doing this research?

I believe that the approach or lens they are looking at this research is the classic “Intermediary Fallacy.” For example, the authors state: ”The author suggests establishing a baseline behavior pattern that can be compared with reactions of various environmental stimuli.” You can cut the animal behavior aspect from the mix but focus primarily on the environmental stimuli. If you have decided to put effort into exploring something that has not been proved, why not cut the intermediary (in this animal behavior) out?

For example, birds are believed to sense earthquakes in advance, sometimes hours. These hours can be critical. Keeping any paranormal, supernatural, and sixth sense aspects aside, it is evident that they can sense changes in the environment, or what the article states as “environmental stimuli.” If we can list the stimuli (ground tilting, groundwater changes, electrical or magnetic field variations), there is no need to study the behavior patterns. Because eventually, we want to understand how the stimuli change before the earthquake.

Think about the effort. Animal experts already have comprehensive documentation of stimuli of birds and animals that supposedly start acting weird before an earthquake. The effort it will take to study how the behavior changes will be less than directly studying how the stimuli parameters start changing before an earthquake. There may also be historical data for some of these stimuli parameters. This approach of focusing on studying animal behavior is what I call the “Intermediary Fallacy.” Animal behavior is just an intermediary. If we already know stimuli parameters, why not study the stimuli directly?

And we get into this trap in many analytics initiatives as well. One of the drivers behind it is frameworks proposed almost half a century ago. We obsess over aspects that do not need to be analyzed. The statistics of 80% of failures of analytics initiatives are not just about failed analytical approaches. It is about a failed problem. definition, which results from a failed analytics strategy. Analytics has been one of the biggest victims of the “Intermediary” trap.