The question of whether a correlation identified in data is spurious or based on a causal relationship – and if so, in which direction – is a well-known problem with interpretation of market data. Answering such questions as close as possible to the underlying real market mechanism is crucial for understanding the analysed domain and for evaluation the consequences of corresponding decisions. We use "empirical causal learning" to identify the types of relationship in observational market data. For this purpose, machine learning methods are used to identify patterns in bivariate data, in which the nature of statistical relationship is known. We then use these patterns to estimate the type of relationship and, if applicable, the direction of causality for the correlations of interest. In addition, we aim to understand how decision-makers handle such situations with unclear type of relationship between relevant quantities. For this purpose, a behavioral experiment in our Decision Lab is also a part of this research project.