Passing the problem to the practice

My previous post posited a number of possible pursuits within a practice, but really failed to bring together any logical focus or even starting point. My interests in this field feel very heterogeneous and underdeveloped - there is a clear difference between believing one has an interest in a particular area, and actually being interested.

I recall a conversation between a couple of friends over fiction writing - they were discussing overcoming problems of plot, where you need to get a character to a particular point or situation in order for the plot to move on. And a piece of advice that was found to be useful was 'give the problem to the character' - find a way to get the character to have to work out how to get themselves into a situation.

Is it possible, in a similar vein, to give the problem of what a practice is to be about, what its concerns are, back to the practice. To find a way within it for it to suggest (perhaps) what it wishes to be about.

This requires from the outset setting up a logic of practice without knowing what that practice is actually concerned with, other than its own self-discovery. This methodology is arguably philosophical in the modern sense, one can imagine it folding in on itself like Descartes meditations until it reaches its Cogito. Or it might never do so, it might continue collapsing, like Descartes practice might have if a certain kind of rigour had been applied to it (like Hume later does)

By what methodology could this be achieved? My first thought was of a database. I was interested that Lev Manovich seem to place databases with the internet as being flat interlink media, contrasting with the hierarchy of a traditional OS. But a database is as much its relationships as its data tables (just as the internet is surely as much its links as its contents). That relationships can be reconfigured does not mean there isn't some kind of implied hierarchy there. The types of possible relationships are defined by the pattern of primary keys after all - if a record has no unique identifier, or its identifier is not referenced, then the record cannot be connecting to others, and possible hierarchies are limited.

The question though is - what would a practice generating program look like to the end user, what would it say, what would it suggest? I can only at this point imagine the most strange and arbitrary configuration, of splicing interests at random with no understanding implied.