[Gzz] The semantic spreadsheet

[Benja Fallenstein]

Have you ever heard about "subjunctive thinking?" Odds are if you have 
you have from me; an old doc about the zaubertrank is third on Google on 
the term. Chris Crawford (2000: *Understanding Interactivity*, highly 
recommended) uses it alluding to "sequential thinking:" Sequential 
thinking is thinking in long chains of arguments, one supporting the 
other; subjunctive thinking, on the other hand, is intercomparing many 
alternative possibilities. You lay out a great number of threads 
(possibilities) converging to one conclusion, paraphrasing Chris.

In our 'bare mind,' we can only hold so many possibilities. With the 
computer, if we can codify rules, we can explore lots of alternatives 
easily. For documents, that's well established-- how does the sentence 
look if I put this here? In spreadsheets, we can try out different 
budgets quickly. Ted has called this 'thinkertoys' I think.

Quoting Chris,

    Let me present an example of subjunctive thinking at work. Suppose
    that you are the manufacturing manager for a high-tech factory,
    and the boss sends you a memo instructing you to reduce your budget
    by 20%. The old, sequential way of responding to that memo
    might look like this:

        Dear Boss:
        Your request [...] is not workable. If I [fire some] workers,
        our output will not keep up with demand. If I cut down on
        electricity, the machines [...] will not work...

    Suppose however, that you simply sent her your budget spreadsheet
    with a note saying, "Please examine this spreadsheet for unnecessary
    expenses. If you experiment with a variety of budget-reducing
    scenarios, you'll quickly see how little fat there is
    in the budget."

    This is a much superior communication. [...]

But the spreadsheet models only the budget, Chris goes on, not the 
actual workflow. Your boss can't try what happens if one machine less 
were used. What's needed, he says, is something like a spreadsheet, 
modelling *that*.

Chris concludes that non-programmers need to learn to program in order 
to make use of these subjunctive capabilities provided by the computer 
(by virtue of being interactive). This was the original motivation for 
designing the zaubertrank, which brought me to Gzz.

But more than programming is needed. You also need a data structure 
capable of representing the intricacies of reality we want to 'think 
about subjunctively.' We need something like a spreadsheet, but capable 
of expressing arbitrarily interconnected things-- potentially including 
social structures. Our bidirectional and orthogonal structures are very 
good at that, or at least much better than traditional structures.

Over these structures, we need to express formulas, so that the computer 
computes some of the connections itself-- like it can compute numbers in 
a spreadsheet. RDF is quite well suited for this, including its existing 
APIs, because of the assumption that some of the triples will be 
generated by logical inference. What I want is a bit simpler: giving the 
computer a set of deterministic formulas for computing the virtual links 
to add to the graph.

(Because links can be to literal values as well as other nodes, this 
includes computing numbers--as in a spreadsheet--as well as computing 
relationships between things, in a single unified mechanism.)

This is what I mean by the 'semantic spreadsheet.' I want to use the 
zaubertrank for entering the formulas and semantic assertions, so that 
everything is expressed in the connections and assembled natural language.

And that's my ultimate vision for Gzz, interlocking with using it for 
taking structured notes and interconnecting applitudes. (Whew!)

- Benja