I use this pattern occasionally, though only when there is sufficient complexity to justify it. I see it as a tool similar to CQRS or hexagonal architecture. It is a way to manage complexity in a more maintainable manner that is easier to reason about and test.

I'll often use it when building out tasks and services, where a simple model (data) is passed into a service (context) and interaction/behaviour specific to that task is injected into the model. In Ruby I'll typically use refinements to do this.

The primary purpose of this approach is avoiding model bloat in my code by being able to decompose additional complexity over time in a flexible way.

If you're interested in getting into DCI more deeply you might want to check out Jim Gay's "Clean Ruby" book (http://clean-ruby.com/)