Re: computer simulation of kinship

Danny Yee (danny@STAFF.CS.SU.OZ.AU)
Sun, 15 Jan 1995 11:09:08 +1000


The problem is not writing a program to represent multi-dimensional
kinship relations; the problem is displaying the resulting graphs.
Your suggestion of rotating a 2D projection of a 3D rendering of
the graph is probably the most practical (though, as suggested by
Mike Lieber, physical models will actually give you real 3D, and are
not impractical for the small graphs you would be dealing with).
Whatever method you chose, you should be warned that when it comes
to displaying graphs there is a lot of arbitrariness in how you
do it -- the graph has a "reality" above any particular graphical
representation of it. (And I suspect that expert system/inductive
learning "representations" of the graph would be more comparable to
participant cognitive structures.)

> I would hate even to begin to reinvent the wheel.

Drawing graphs "nicely" is a common problem, and has definitely been
tackled before (though I don't have any references handy).

> I have
> been discussing the matter with a colleague in Computer Science here,

Who's probably already explained what I just wrote, but hey! long live
inter-disciplinary cooperation!

Danny Yee.

P.S. For small graphs (and kinship graphs aren't going to be big by
computing standards) you could probably do what you want pretty easily
with any ordinary math package that supports graphics. I know
Mathematica could do it, for example.