Is the PIE past its sell-by date?


Alan Dix
Lancaster University

Talk given at 20 Years of HCI at York : Kings Manor, York, 4th June 2004

Full reference:
A. Dix (2004). Is the PIE past its sell-by date? Talk given at 20 Years of HCI at York, Kings Manor, York, 4th June 2004


download slides of talk (PDF, 2.6M)
Alan's pages on formal methods in HCI
Upside down As and algorithms - computational formalisms and theory In HCI Models, Theories, and Frameworks: Toward an Multidisciplinary Science. John Carroll (ed.), Morgan Kaufman, 2003. pp. 381-429
Formal Methods for Interactive Systems. Academic Press

send a cracker

third edition out now!!



download slides (PDF, 2.6M)

outline ...

20 years on - the PIE and all that
I started to work in HCI when I joined the "5 man project" in 1985. Part of our aim was to formalise Harold's GUEPs (Generative Usability Engineering Principles) and this lead to the development of generic models of classes of interactive system. Unlike specifications of particular systems these were designed in order to aid understanding of general issues. The PIE model was the most generic of these taking a minimal view of single-person interaction: input, output and processing. Despite its simplicity, the PIE model enabled us to descriptive properties such as WYSIWYG (what you see is what you get) and also to proof things about undo.
There were numerous variants of the PIE model and related models, perhaps the most complex being 'the cube', developed with Roberta Mancini. This modelled 'layered' systems where an underlying system had some added higher-level functionality, allowing is, in particular, to prove more detailed properties of undo and also to model 'back' and history features in browsers.
Although this is not as central to my work now as it once was, this kind of modelling has not ceased. As a way of measuring the continued interest in this sort of formal modelling I looked at citations in citeseer to my first book (Formal Methods for Interactive Systems). Very much to my surprise I found that this has remained relatively stable over the period citeseer covered (1993-2001). Recently too I wrote a chapter of formalism in HCI for Jack Carroll's theory collection and of course there are several chapters relating to formal methods in a very successful HCI textbook :-) This type of formalism is particularly popular in Italy although the Italian translation perversely has skipped these chapters!
further back - a formal methods success story
Strangely my greatest success story in applying formal methods to HCI came before I went to York! In 1984 I was working in Cumbria County Council in the Data processing Department and building transaction processing (TP) systems. Like web-based systems these have a single server with many terminals submitting screen-forms. In order to make sense of the increasingly complex programs I used flowchart, not of the program, but of the human-computer dialogue. At that stage I did not know the words, but I was using a form of formal dialogue specification. This lead to an enormous productivity gain, but it is only now, in retrospect, I can see why it was so powerful.
formal futures - ubiquity and physicality
The PIE model was about single-user glass screen , keyboard and mouse interaction. Extensions dealt with multi-user interaction and other topics, but still largely restricted to standard GUI-style interaction. However, now computers are permeating every aspect of our lives both at work and at home. Traditional interfaces tended to be quite broad and shallow in terms of their semantics. For formal specification this meant that the main advantage of formalism was definitional - helping clarify understanding. For a mathematician, proofs and analysis were shallow or trivial. However ubiquitous systems often consist of relatively simple individual components interacting in relatively complex ways ... just the sort of thing formalism and mathematics are good at! So formal methods are more important now than ever.
Working with Masitah Ghazali at Lancaster we are looking 'fluidity' the way in which the physicality of simple knobs, buttons etc. is exploited in everyday consumer electronics. The aim is to mine the rich knowledge embodied in these existing artefacts in order to build transferable design principles for novel tangible devices. As part of this process we build state diagrams of the physical device and corresponding state diagrams of the logical functions they control and relate the two.
Another aspect of ubiquitous interaction is where sensor-based systems do things 'for you' without bidding. A common example of this is car courtesy lights which switch themselves on and off when you unlock the car, open doors, start/stop the engine etc. I call this incidental interaction, although there are also several other closely related terms in the literature. Given the relationship between sensors and user activity is uncertain, design methods must be robust to this uncertainty. This use of sensors and intelligence within a forgiving interactional context I call 'appropriate intelligence'. As a first step towards a design methodology with students we have been looking at scenarios, annotating them with varying levels of desirability of controlled state (e.g. interior light in car), and then using models of sensor state in order to match these.

In summary, I conclude that 20 years on ... the PIE is still fresh ;-)

references ...

The PIE model


The Cube


Recent chapters ...


Success story


Formalism in ubiquity


Alan Dix 6/6/2004