Xref: utzoo comp.ai:1571 comp.graphics:2375
Path: utzoo!mnetor!uunet!mcvax!ukc!strath-cs!glasgow!gilbert
From: gilbert@cs.glasgow.ac.uk (Gilbert Cockton)
Newsgroups: comp.ai,comp.graphics
Subject: Re: expert systems for graphic design?
Message-ID: <1027@crete.cs.glasgow.ac.uk>
Date: 26 Apr 88 09:45:09 GMT
References: <10494@sunybcs.UUCP>
Reply-To: gilbert@cs.glasgow.ac.uk (Gilbert\ Cockton)
Organization: Comp Sci, Glasgow Univ, Scotland
Lines: 62
Keywords: design, graphics, maps, cartography

In article <10494@sunybcs.UUCP> dmark@sunybcs.UUCP (David Mark) writes:
>REQUEST:  Does anyone our there know of any expert systems (or other
>kinds of software systems) for evaluating the graphic design of a
>display?
>If people reply to me via email, I will summarize responses to the net.
Sorry, email can be flakey from Europe.

Main references I know are

%A D.J. Streveler
%A A.I. Wasserman
%T Quantitative Measures of the Spatial Properties of Screen Designs
%J INTERACT'84
%V 1
%I Elsevier/IEE
%P 124-133 (participants edition)
%D 1984
%O 1985 edition pub. North Holland

%A T. Tullis
%T Designing a menu-based interface to an operating system
%J CHI '85
%P 79-84
%D 1985

%A T.S. Tullis
%T Optimising the Usability of Computer-Generated Displays
%B People and Computers: Designing for Usability
%E M.D. Harrison and A. Monk
%I Cambridge University Press
%C Cambridge
%P 604-613
%D 1986

These measures covered are useful, but very crude.  Graphic designers
are not ones for writing things down, nor can I see them rushing to
have their knowledge elicited by production rule hackers. It is far
more efficient to find a NUMBER of graphic designers locally and ask
them to evaluate your display layouts.  Then use your judgement to
decide on which advice to take.

If you've got a long time, you could implement some alternative
designs for aspects of the display and get a human factors expert (not
a theoretical psychologist) to help you to compare human performance
effects of the design alternatives.  You could also discuss design
alternatives in the first place with such an expert.  Use paper here,
as it's more efficient in early design than most screen generators -
MacDraw is a good early prototyping tool.

Finally, try things out on representative end-users, who may like
neither the aesthetics of your preferred designer nor the optimum
performance of the human factors experiment.

This all seems harder than just running a program, but I can assure
you that it is all a lot easier than trying to design one to do an
equivalent job.  We do not have a computational account of good
graphical design, are unlikely to gain one in the near future, and
probably never will reduce aesthetics to some thing as ugly as a
Turing equivalent formalism.  Of course, the phenomenological aspects
of end-user preferences can NEVER be automated or simulated, because
such phenomenology is defined to be human experiences and
categorically absolutely nothing else. It is as computable as a daffodil!