• Are we sure that meta OCL exists?
• If meta OCL exists, is it a good thing to learn by a neural network?

• What makes meta OCL a good thing to learn for a neural network?

• How do the authors find out whether meta OCL is exhibited by neural networks?

• Why do the authors say neural networks as a general term instead of a specific model they test? Is there anything to suggest that it is a general phenomena?

• Is there a faster way to demonstrate whether Meta OCL exists?

• Can the way to demonstrate meta OCL be adapted to demonstrate existence of other phenomena?

• Can the detection process be made more accurate?

• Can this be made into an anytime algorithm?

• Can we make this into a streaming algorithm.

• What other data types will it work with? What about uncertain data?

• Why does this work?

• Can this detection method be made to work on a distributed system?

• How to apply this to a novel setting?

• What are the assumptions being made by the paper?

• What are the main parameters?

• How can it be made simpler?