Skip to main content Skip to navigation

Department of Computer Science

General Feedback

You can find the entire list of abstracts and modelling study proposals submitted on a separate page. Because of changes to the protocol used for collecting responses using web forms, these abstracts are anonymous, though of course in point of fact I do know the authors of several submissions because of previous discussions we have had. Please find your abstract, and my comments, and consult me if you require any further clarification.

First some general comments on the submissions. Note that you are expected to motivate and complement your modelling study with a theme. There are numerous examples of papers with associated models in the list of EM papers, and some of these have been the subject of lectures. (Consider for instance the OXO laboratory and paper #033, or the ant navigation model and paper #110.) It is very important to have a title that conveys a sense of theme, rather than just a plausible name for a model. It is also important to ensure that your abstract doesn't lapse into a description of the model to be constructed, or the mode of its construction. Finally, take great care to check spelling and grammar, and be sure to use "Empirical Modelling" rather than "empirical modelling", which has an accepted meaning but means something different. One noticeable feature of the submissions was the emphasis given to modelling studies in your proposed weightings - bear in mind that modelling may prove more difficult than you think, and try to develop ideas about how to strengthen the paper component of your submission as you tackle your models.

Categories of submission for WEB-EM-7

In reviewing the submissions to WEB-EM-7, some general categories of modelling study came to mind. They are as follows:

  • A: Games with a discrete / combinatorial flavour
  • B: Traffic related simulations
  • C: Newtonian mechanics related simulations
  • D: Algorithms and data structures related
  • E: Games with an analogue flavour
  • F: Engineering / science scenarios
  • G: Computing systems scenarios
  • H: General simulation
  • I: Mathematics learning
  • J: EM evaluation and critique
  • K: Visualisation

I have appended some individual feedback and comments on your submission as it appears in the listing - I have done this in a public way so that everyone can benefit from reading this.

For each submission, I have indicated which of the above categories are most relevant to your proposal. I also append some general reflections on each of these categories together with some notes on resources - these may be helpful to you in preparing your final submission. In virtually every case, I have been able to identify good reasons for considering the use of Cadence / Cadence-with-EDEN. It is only fair to reiterate however that Cadence is a research prototype, that there will only be very limited support for debugging, and - whilst we shall be delighted to receive submissions that contribute to Cadence research - it will be perfectly to gain an excellent mark without making use of Cadence. If you do make use of Cadence, you strongly advised to adhere closely to the patterns of use illustrated in the lab sessions, and to document any problems you encounter - and solutions you develop - using the Cadence DIY thread in the CS405 forum.

A: As the lectures have illustrated, this area is a particularly rich source of EM studies (cf. the OXO model, the WUMPUS model, the Sudoku models, wordgame lab, etc). One of the features of such games that makes EM appropriate is that it is possible to play with the rules and change the context in playful ways that are totally within the spirit of the application. It is a good idea to exploit this flexibility about meaning in your approach, and to avoid the trap of making a model that is directed too strictly at the goal of "implementing the standard game" (though you can make such an implementation without diverging from EM principles in the development). One of the most interesting aspects for future EM work in this area is to tackle the other aspect of game-playing where imagination and creativity are important - in playing and analysing a game without tinkering with the rules. The WUMPUS model has this flavour of a meta-agenda. Note that some standard activities, such as the use of minimax strategies for evaluating moves have yet to be successfully implemented in EM, despite the fact that EM may be well-suited to enhancing them to develop richer strategies. One obstacle to this has been the difficulty of representing many different possible positions in a game in the setting of an EDEN model. This should not be such an issue if manipulation of observational contexts as in Cadence is used effectively, but this has yet to be done by anyone.

B: There have been many EM modelling studies relating to traffic simulation (an early effort can be found in trafficlightMendis1997). A significant issue is that traffic lends itself to multi-agent simulation, whereas EM on the EDEN platform tends to be best oriented towards modelling individual agent perceptions (cf. the ant navigation). This means that any multi-agent modelling (e.g. such as Chris Martin's studies of avoidance strategies used by people meeting in corridors) has been confined to a modest number of agents (in Martin's case at most 20 people) and very simple visualisation (person is represented by a circle, and their gaze by a line etc). It will be difficult to make a significant breakthrough in this area without exploiting Cadence, and the scope for cloning agents etc. One of the WEB-EM-6 submissions did some useful preliminary work in this direction.

C: Mechanics and systems of bodies have always been a popular choice of topic for EM students. Andy MacDonald's 3D-room model is one example of modelling with physical laws based on Newtonian principles. The kind of integration mechanism required to calculate the values of dynamic observables (e.g. forces, accelerations, velocities, positions) in EDEN was illustrated in Ian Bridge's vehicle cruise control simulation. It seems plausible that other implementations can be made using edenclocks, a feature introduced to EDEN only relatively recently, but I don't know of a good precedent for this. The potential for using Cadence here is obvious: its process-like observables lend themselves to interpretation as dynamic quantities. The use of "willbe" definitions is clearly highly relevant here, but it has become apparent that making dynamic models using only such definitions is fraught with synchronisation problems. The introduction of the "is" definition into Cadence should resolve these problems, and we are expecting to see progress in this direction in the next generation of Cadence models.

D: The EM heapsort modelling study illustrates the potential for linking EM to the conventional implementation of data structures and algorithms. It is important to recognise that the formal notion of an algorithm is predicated on there being some functional goal for which the algorithm supplies a generic solution. A good algorithm is typically optimised towards achieving its goal, and "optimisation for purpose" is an activity very different in character to "making sense" and "gaining understanding" prior to establishing a specific goal. For this reason, the efficient implementation of algorithms in EDEN is characteristically associated with the procedural ("traditional") rather than the definitive ("EM") aspect of the hybrid environment. As when applying EM to games (see A), there may however be scope to apply EM principles at the meta-level in the algorithm domain e.g. to explore issues about deployment and performance of algorithms. The use of Cadence may also offer an alternative perspective: within Cadence it is possible to specify process-like relationships using "willbe" dependencies, and this makes it possible not only to specify functional relationships in a non-procedural "data flow" idiom, but also to elaborate them in an EM style. (This is one of the modes of use of Cadence that is represented in sophisticated geometric modelling within the Stargate model.)

E: Other varieties of games that have been studied in EM include traditional two player and team games. Cricket simulation has been a topic of study in many previous projects (see e.g. Zhan En Chan's account of the Cricket Project, introduced as a exercise in software development in the second year early in the 1990s). Other games featured in EM case studies include football, billiards, basketball and American Football. Modelling these games can combine cognitive concerns with elements of simulation of the physical environment. Even "very simple" activities represented in such games (such as catching a cricket ball) raise challenging issues where the identification of agency, observables and dependencies is concerned. This is another area in which the blend of analogue process-like activities (such as the motion of a ball) and discrete decision-making processes reflecting human agency motivates the introduction of Cadence.

F: One of the characteristic features of many engineering and science scenarios is the presence of physical devices or experimental apparatus. Making EM construals for such physical artefacts is an excellent focus for a modelling study. The term Interactive Situation Model (ISM) can often be applied to EM construals of this nature, and there are a number of papers and reports that adopt this term. Examples of artefacts that have been modelled in the past include digital and analogue watches (digitalwatchFischer1999), central heating systems and a vehicle cruise control system (cruisecontrolBridge1991). Whereas a traditional model of a physical system is liable to be constrained to have a specific functionality and mode of interaction and interpretation, your objective in EM should be to expose the pattern of observables, dependency and agency that are involved in a construal of the system, and so open up the possibility of richer forms of exploratory and experimental interaction. The distinction between traditional "theory-based" models and construals is the subject of Beynon and Russ's paper Experimenting with Computing.

G: There have been a number of EM modelling studies directed at computing systems. They include a number of previous WEB-EM submissions. In simulation using EDEN, it is impossible to make models that incorporate more than a modest amount of sample data, and this has meant that applications have typically been directed at educational goals. As is illustrated to some extent with the vimodesBeynon2006 model, there is considerable scope for using visualisation techniques to make otherwise hidden state explicit and this principle may be a good way in which to provide facilities to help the users of software utilities etc. There are much better prospects for constructing more representative models of computer system operation using Cadence. The fact that DOSTE was first conceived as an alternative operating system indicates that it may even be possible in some instances to blend realistic simulation of functionality with an educational role.

H: Simulation using EM is of most immediate interest when it engages with the possibility of alternative construals. Where there is a standard construal that is unlikely to be improved by the modeller, simulation using EDEN is likely to be much inferior to what can be done by specially developed tools. The benefit of an EM simulation lies in the possibility of expressing non-standard construals, unusual/unfamiliar referents and considering unexpected contexts. When you construct an EM simulation, it is advisable to demonstrate these qualities. Making a standard model in which you have privileges to change a few observables in well-understood ways may be a good way to demonstrate dependency in action, but the true significance of a dependency is only appreciated when the possibilities for redefining it are considered. As in many other modelling contexts, the size and scope of the models you build in EDEN is quite limited, and it is a good idea to consider using Cadence to address these limitations.

I: Learning mathematics has been a popular topic for EM construals. Even elementary topics, such as learning about different number bases, have been the subject of good case studies. There is plenty of scope in this area to draw on the thinking that we have done on EM for educational technology. Studies of learners interacting with EM construals would be particularly welcome: as experience with the Sudoku model has shown, this interaction has some unusual qualities where interactive evaluation is concerned. If you use EDEN, the most appropriate topics tend to be those that involve discrete structures - it is not particularly easy to generate graphs of continuous functions etc in EDEN. The use of Cadence is a promising way to overcome these limitations.

J: Some of the most interesting developments in EM (including the development of Cadence as an alternative EM tool) have been initiated through critiquing previous thinking and practice. It has often proved difficult to understand the relationship between EM and other perspectives initially, and it is unreasonable to expect students familiar with EM only through CS405 to be able to make a fully informed critique. Provided that you are able to demonstrate sufficient understanding of EM and can motivate your submission suitably, it is quite acceptable to build models using other tools or to focus primarily on making commentaries on existing EM models and papers. Contributions that help to improve our understanding of Cadence-with-EDEN as an EM tool are particularly topical.

K: Visualisation is a fundamental theme in EM because of the importance of 'rooting all knowing in relationships that are given in experience'. There is scope to do more than to give visual expression to basic components of physical systems though: other forms of visualisation of data can also play a crucial role in developing construals. One example of this is the Attribute Explorer first developed by Bob Spence, and later adopted by IBM (cf. attributeexplorerRoe2000), in which representations of data are developed by exploiting subtle forms of dependency. EDEN is able to cope with small data sets, but in typical applications it would be useful to apply similar principles to much larger volumes of data. There is scope to explore the use of Cadence both in handling more data and in representing the range of observational contexts in clearer and more flexible ways.