Automated Causal Detection

Estimand has just completed the anticipated update to its causal detection model. Named CR-2, this model represents the first automated causal detection system. Without human involvement, CR-2 can discover causal relationships among time series datasets.

Perhaps less sensational than many of the AI models being developed, CR-2 represents the first step toward Artificial General Intelligence (AGI). The foundation for a machine that understands the world in which it operates.

As Turing Award winner Judea Pearl warns, “Deep learning has instead given us machines with truly impressive abilities but no intelligence.” Large Language Models (LLMs) can determine the next word in a sentence but cannot understand what it means. In contrast, CR-2 is a machine with understanding and a more subtle use case.

Correlation vs. Causal

Correlation is a measure of how much two variables are related. In contrast, causation means that one variable causes another. Known statistical calculations determine correlation. Causation is only proven through experimentation, often with subject matter experts.

A Spearman Correlation measures the monotonic relationship between two variables [Ramzai 2020]. This means that when one variable changes, so does the other. It may be a positive change in one that produces a negative change in the other variable, or it can be a positive change in one that creates a positive difference in the other variable.

However, the relationship measured by a Spearman Correlation is not linear. Meaning it is not at a constant rate. Therefore, it does not prove or disprove causation. Instead, it often signifies a relationship between the variables and another unknown variable.

For example, whenever my roof is wet, my neighbor’s is wet, too. There is not a causal relationship between the two variables. Instead, a correlation between the two variables points to an unknown source of the phenomenon — most likely rain.

Causal Detection

Recently, my wife’s little dog was experiencing bright red skin, dampness, hair loss, and a smell similar to a sweaty gym sock. He was scratching, constantly liking his paws, and appeared very uncomfortable.