H. Vanoostendorp et Bj. Walbeehm, TOWARDS MODELING EXPLORATORY LEARNING IN THE CONTEXT OF DIRECT MANIPULATION INTERFACES, Interacting with computers, 7(1), 1995, pp. 3-24
Citations number
35
Categorie Soggetti
Computer Sciences, Special Topics","Computer Science Interdisciplinary Applications
The characteristics of direct manipulation interfaces (DMIs) are exami
ned. The main purpose of this examination is to provide ideas for futu
re research on modelling exploratory learning in the context of using
DMIs. Four topics are discussed: the perceptual characteristics of DMI
s, exploratory learning and display-based problem-solving in general,
modelling human-computer interaction in the context of DMIs, and the c
onsequences of DMIs for modelling the interaction by means of a produc
tion system. Specifically, the questions that are discussed are: first
, how do DMIs afford, encourage and support exploratory learning, and
how can typical DMI characteristics such as the objects on the screen
be included in models of user behaviour? Second, what are the characte
ristics of problem-solving and exploratory learning in the context of
visual displays? Third, how is novice behaviour and, more generally, p
roblem-solving modelled in the context of human-computer interaction?
In the final section, suggestions are made based on the topics discuss
ed, with the aim of presenting some steps towards developing a model c
onsisting of production rules that can simulate human interaction with
DMIs more adequately than has been the case thus far. Two important c
onsequences of DMIs for modelling human interaction are discussed. Fir
st, the external display of DMIs allows recognition instead of recall.
Consequently, production rules can be more recognition-based. Second,
with regard to the structure of production systems, the mechanism of
partial matching is proposed to account for errors during performance.
Constraints and affordances can be accounted for by proposing product
ion rules to fire context-dependently, and by assuming that production
rules can be meaningfully grouped and actively scanned for a match.