Comments for Nowviskie Paper
Ludic Algorithms

Devon and I read and discussed Beth's paper and I put our comments and questions together to present to you.  Devon will jump in if I miss something.  We independently had the idea of putting some of our conversation into Llullian form, hence the volvelles.

Beth's paper is about the work (or the 'Arts') of the "medieval polymath and sometime poet, rake and martyr named" Ramon Llull.

Llull invented a series of systems that allowed their users to recombine discrete symbols in mechanical ways. Llull's work is now often misunderstood as purely mechanical, and thus unlikely to lead to anything of interest. Beth argues, however, that Llull's various arts were intented to stimulate reflection and interpretation by their human user (the 'artista').

Properly understood, Llull's work raises many questions for "humanists with fresh ideas about the relation of mechanism to interpretation."

One of the things we liked about this paper is the way that it opens out in so many directions, exemplifying a playful and combinatoric mode. Beth approaches Llull's work from many directions: from literary studies of satire, internalist histories of computer science, meta-level discussions of digital humanities, cognitive science, and so on.

Raises many questions...

What are, or should be, the scholarly primitives for the digital humanities? What balance of deductive and inductive reasoning should our tools favour? Can they help us with what CS Peirce called abductive reasoning?

Can the mistakes that a machine makes ever be interesting? Lady Ada thought not, but she didn't have any experience with actual computers. Alan Turing and Grace Hopper, who did, knew that computers often surprise you. Grace Hopper has a nice description somewhere about how she originally thought computer programs would just work, and what a huge surprise it was for her to discover that she was going to spend the rest of her life trying to root out bugs. 

(ASIDE: As another example, I was once talking to a machine vision researcher who told me about visiting Japan in the 1980s and seeing some of their brand-new facial recognition software in action.  For some reason the system kept recognizing some of the Americans as if their heads were upside down.  Eventually the embarassed Japanese figured out that their system hadn't been trained on bearded men with male-pattern baldness.  So the system assumed that the men's beards were actually on the tops of their heads.)

In the paper, you define a ludic algorithm as

"a constrained, generative design situation, opening itself up--through performance by a subjective, interpretive agent--to participation, dialogue, inquiry, and play within its presecribed and proscriptive 'computational spaces'. ... the desired outcome of a ludic algorithm is sheer, performative and constructive enactment of the hermeneutic circle, the iterative 'designerly' process we go through in triumphing over interpretive or creative problems we pose ourselves."

This seemed to us to offer many interesting connections with other papers in the volume. How does the ludic algorithm relate to the distinction between serious and casual games, raised by Josh and others?  How does it relate to the distinction between playing a game and playing *with* a toy?  Rob's 'barely games' are ludic, but could they be algorithmic?  Could the outcome of a ludic algorithm be screwmeneutical enlightenment?  Lots to discuss.

When we build a constrained system, how do we signal those constraints to the user or interpreter? How do we allow the user to relax the constraints or put others in place?

There are a few unexplored directions that we're curious about.  Given Llull's Islamic context, what is the relationship between his volvelles and the astrolabe, the quintessential analog computer of the day? We know that Llull himself knew some Arabic and hoped to convert Muslims to Christianity.  Another computational technology that was taking off in Llull's world at the time was the mechanical clock, which also had religious implications (e.g., waking Christians to pray at specific times).  I'm not sure what role astrolabes and clocks play in internalist histories of computer science ... I've usually found them in more general histories of science and technology, or the history of astronomy.

You may already know this, but if not, I'm sure there are very interesting connections to be made between Llull's work and the anthropological literature on randomness and divination.  Many scholars have studied divinatory practices as a kind of a conversation between a series of essentially random events and a human interpreter (or interpretive community).  In a relatively stark form, one of the ethnomethodologists (I forget which, but could track down the reference for you) set up an experiment where he had undergrads type questions into a terminal, and supplied each with a random YES or NO answer. He found that the students tended to reframe questions and interpret the overall conversation so that it would make sense, even when the individual answers were contradictory, in fact nonsensical.  Here is a made up example:

- Do you think I should get married?
NO
- So I shouldn't have a relationship with X?
YES
- So I should postpone my wedding?
NO
- I guess I should cancel it?
NO
- So we should move in together?
YES

We're pretty good at making sense from nonsense, at filling in more than our half of a conversation.  It helps us read one another's minds and get along.  If we treat a human baby as a conversational partner, they eventually become one.  We extend the same courtesy to pets, to artifacts, to random events.

In your discussion of algorithms, you raise the problem of underspecification.  How do you know when you've defined your terms clearly enough so that they can be implemented by machine?  A similar conundrum plays a role in the history of science, where it is known as "the experimenter's regress."  When a student replicates Galileo's inclined plane experiment, they know what value they are supposed to get for the acceleration of gravity.  But how did Galileo know?  If you are the first one to do an experiment, you don't know what the results are supposed to be.  If you replicate it, you may only be replicating your own errors.  The reason we bring this up is because "an experiment" and "a computation" are deeply similar.  You set up some physical conditions and run them out.  If you don't know what to expect, you are performing an experiment.  When you do know what to expect, you're performing a computation.  So there might be something interesting to look at there.

Is the underspecification of algorithms a problem for programmers or computer scientists?  Not as much as you might think, as long as you take the view that programs are primarily a way for human beings to communicate novel ideas about methodology with one another.  They are only secondarily designed to get machines to do something. (That formulation comes from the Wizard Book, an infamous LISP textbook).

Alternately, we might ask about the role of instrumentality in humanities research. Can we  borrow, adopt or adapt the insights learned from research on, say, scientific instruments to cultural instruments? For an example of this, we were thinking of something like Baird's "Thing Knowledge." 

How big is the divide between scholar-user and scholar-designer? Personally, I would argue that mechanically-assisted algorithmic methodology *can* be an interpretive strategy, but that it will be most successful for humanists who can do their own programming.  It's an old argument in the digital humanities: humanists used to have to be able to read and write many languages (like Latin, Greek or Arabic); now they need to be able to read and write computer programs.  It is difficult to set that disciplinary shift in motion, however, without having faculty who already know how to program--or at least are committed to learning how. Every year my public history MA students tell me that they will just "hire a programmer," and I have to explain some sad facts about relative wages.  If we can't write our own programs, we're pretty much stuck with unsuitable and general purpose tools.

Speaking now for myself, there are a few things that I would change if this were my own paper.  In particular, I don't love the section on cognitive science.  It seems to take their account of their own work and 'discoveries' at face value, rather than treating them skeptically.  Briefly, cognitive scientists believe that thought is computational--and maybe emotion, and maybe even conscious experience, too.  They believe that "the mind is what the brain does."  They may be right, but I think there are a number of critiques that they haven't fully satisfied.  Some of these come from phenomenological or hermeneutic approaches (like the Winograd and Flores book you cite), some come from people like Maturana and Varela (who you also cite), some come from ethnographers like Lucy Suchman, some come from environmental or situated cognition. And so on.

We'll leave it there, and let other people jump in.