Deriving exact causal relations between two variables is a difficult task due to the possible incompleteness of the data and the context. It is difficult to identify whether among two correlated variables there is a causal relation, which variable is the cause and which one effect. Also, another possibility is that there exists an external variable not present in the dataset, which has the same cause-effect relationship for both the given variables being examined. Without any contextual information, it is not possible to determine the cause-effect relationship between two variables.
For example, the recent COVID19 has multiple symptoms like cough, fever etc. But it cannot be inferred that a person is suffering from COVID19 just because he/she is suffering from cough and cold. There is something missing in this relation. If the person has recently travelled to/from the countries affected, and the person is sick, then it can be inferred that he/she might be suffering from Coronavirus. Clearly, the addition of a new variable “Recent travel” in the data is capable of changing the way people might interpret the data.
Our goal is to come up with a web application that provides the users with an interface to incorporate such user-defined additions/changes to the dataset such that the final result makes more sense in terms of real-world data. This web application will allow the user to get an initial causal model and verify its credibility using personal intellect and further suggest the possibility of new latent variables that would complete the causal model with valid causal relations. We aim to extract such relations and provide the users with a game like interface, using which they can put their domain knowledge and extract relevant data from the web which can be fed to the tool, to identify the actual causal relations. If the data is incomplete, the facts derived may not be representative of the actual causal relation between certain data fields. For example, a dataset consisting of only 2 fields: ‘SAT score’ and ‘Income’ might result in a causal model which suggests that high income leads to high SAT score. This is not entirely true, since there might be various other additional factors that actually contribute to getting a high SAT score. The application will allow the user to look out for these variables, which might be missing from the current dataset but might be available in some other datasets, which can be added to build a credible causal model.