Papers

Stevens C., Visscher B., Addis T.R.

2006 IEEE International Conference on Systems, Man and Cybernetics

ABSTRACT: We present a novel approach to handle uncertainty, by assuming an irrational view of the world thus breaking with the tradition of taking a rational view. Changing this view allows us to think differently about the nature of uncertainty. Based on this new understanding, we propose a system that can cope with an imprecise, uncertain and changing ‘world’. The proposed solution handles irrationalities, which includes uncertainty, by generating different hypotheses of the environment and testing these for their validity thereby fully automating the abductive loop. Since the proposed system makes no assumption that any knowledge about the world will ever be completely reliable, all knowledge is susceptible to change. With the full automation of hypothesis generation, inference mechanism and validation, there is no apparent problem domain with which the proposed system could not cope.

Visscher B.

PhD Thesis, University of Portsmouth

ABSTRACT: Software is sometimes considered the most flexible part of any system. The flexibility however appears to exist only if the software was created ‘correctly’. When software is not appropriately adaptable in relation to the problem domain, the cost of modifications increases over the software’s lifetime until it eventually becomes too costly to maintain. To reduce the cost of maintenance several solutions have been proposed either by improving the programming language, design method (methodology), requirements analysis or others. To understand the problem of maintaining flexible software and the increasing complexity when software is not appropriately adaptable, this thesis presents an empirical investigation into software structures in an attempt to determine the influence that language, development method, team organisation have had on the programs analysed. The investigation showed, surprisingly, that any influence of the design method, programming language, team organisation, development environment, problem domain and program purpose is statistically insignificant for all the analysed projects. The evidence shows that the only statistically significant factor influencing the structure of programs are what appear to be cognitive preferences or cognitive limitations. The cognitive organisation that results from these cognitive preferences and limitations is similar for all analysed programs......The realisation that the structure in which changes are implemented needs to remain close to the psychological organisation to order to handle complexity means that software is not infinitely flexible. The limitation in software flexibility is not a formal one but a cognitive one. Although the problem of achieving flexible software cannot be ‘solved’, the best way to handle it appears to be by somehow increasing the upper boundary of cognitive limitation by using, for example, automatic inference mechanisms over the program or, more drastically, by replacing the limiting factor by something less limited.

Addis T.R., Gillett R.

BCS SGAI AI, App. and Innovations in intelligent systems XIII, ISBN:10-184628-223 3 p137-150

ABSTRACT: WISE Expert is a general-purpose system that can be used for monitoring or controlling, in real time, complex systems that have recurring sub-structures. The system has been developed using a unique schematic development tool that ensures coherency of structure during design and construction. The design of the Expert System takes advantage of a distinction between the monitored system structure and expert knowledge so that the structure description can be used to generate specific rules for the system automatically. The system has been tested as an overseer during the running of trainee mariner exercises with a liquid cargo simulator and is now operational at over 35 customer sites throughout the world.

Addis T.R., Visscher B., Billinge D., Gooding D.C.

Grand Challenges in Computer Science, University of York

ABSTRACT: In response to the grand challenge for computer science during a workshop held in Edinburgh in November 2002 we identified an essential problem in computing that has yet to be addressed directly, see Appendix A for UKCRC Grand Challenge criteria (Addis et al., 2004). We show that this problem, originally identified by Wittgenstein circa 1945 (pub. posthumously, revised translation 1966), explains a barrier that prevents people from communicating seamlessly with computer systems. It explains many of the apparently insoluble problems that beset human computer usage such as; context dependency in natural language understanding, the generalisation problem in machine learning/neural networks and effective data retrieval. Although many solutions have been attempted, nobody has yet directly addressed the underlying cause. This cause can be characterised by the notion of the ubiquitous existence of irrational sets that emerge from the fundamental nature of human language and a continually changing knowledge of the world. We suggest two possible routes to solving this problem.

Addis T.R., Visscher B.

Unpublished

ABSTRACT: An extended functional dependency analysis was developed in order to investigate the natural structures of programs (if it exists). This initial study was conducted with the eventual purpose of exploring the effects of programming language and development methods on program architecture. The analysis was designed to produce a uniform representation across different programming languages. Further, this representation was chosen so that it merges the classical distinction between data and procedure into a single functional form by using an extension of the normal functional dependency, as used in functional analysis, to include functional behaviour. Using this extended functional dependency analysis, several programs were transformed from two language sources and some natural bounds were discovered that seem to be independent of both language and method. Based on these natural bounds, two additional analyses were developed. Both these additional analyses, the function role analysis and keystone analysis, exposed further constraints that could also not be explained by language, method, architectural or problem domain constraints alone.

Addis T.R, Townsend Addis J.J.

International Journal of Human Computer Studies, Vol 56, No 4, pp331-422 , ISSN 1071 5819

ABSTRACT: Clarity is a functional schematic programming language currently freely available to the community. It is a programming environment that allows a user to draw a program as a set of directed graphs. The term schematic is drawn from the traditions of engineering where the diagrams that represent electronic circuits or those of physical objects are often referred to as schematic drawings. A schema is a set of pictures or graphs that represent a program or working model. A schematic is taken as a system of tokens and structuring rules that expresses a program, model or concept; it is a graphical language. This paper introduces the principles behind design and issues to be considered when dealing with complex systems. The reasons why a 'functional' representation provides a non-invasive approach to design and forms the basis of 'good' design are described. In particular, the advantages of using diagrams is shown to be because the schema constructions make the structure of complex systems explicit as well as make a functional representation more intelligible than its sentential equivalent.

Addis T.R, Townsend Addis J.J.

International Journal of Visual Languages & Computers, Vol. 12, pp 689-715, ISSN 1045 926X

ABSTRACT: Designers in general have used diagrams and sketches to help in the process of creation. This is particularly so for system designers whose output is a set of programs. It would seem reasonable that the conversion of diagrams directly into a program would be desirable and yet the work of Green and Petre [12,13,14,19] and Citrin [8] has placed doubt on the viability of graphical programming notations. Some of this work is reviewed in this paper. The use of secondary notation and the match-mismatch hypothesis is reconsidered in the light of functional programming. It is proposed that much of the criticism of graphical notation is due to the imperative (or process orientated) nature of programming. Many of the limitations observed in using graphical notation are lifted when functional programming is used. Eight engineering dimensions and four engineering relationships (coherence's) are proposed to describe programming environments (including notation). The key to the success of a functional language as a general representation as well as its coherence with a graphical notation comes from its unique extensibility. Support for these arguments is drawn from examples of a schematic programming language used for industrial scale projects.

Gillett R., Addis T.R

GasTech2000, Houston, Texas, USA

ABSTRACT: For a ship to be run in a safe, efficient and cost effective manner requires a balanced combination of the right equipment and control systems for the tasks is expected to perform together with a sufficient number of personnel who are skilled in the relevant areas.  In recent years this balance is becoming increasingly more difficult to achieve not because of problems with the equipment, which is constantly being improved, but because of a lack of availability of personnel with the required skills.  We describe how a full simulation of a ship’s cargo handling system in conjunction with an intelligent system can provide a self contained training system on a laptop. This amalgamation can provide expert training for mariners without the need of a shore-based course. In the control system context, by incorporating the knowledge of many experts through the use of Intelligent Software, the total required knowledge needed for the safe and efficient operation of the ship or installation can be maintained.

Addis T.R, Gooding D.C.

AISB Symposium, 6-9 April, University of Edinburgh. Pages 19-28, ISBN 1 902956 04 4

ABSTRACT: Computer simulations of learning, discovery and problem-solving generally neglect the social aspect of these processes. EXPLORE v. 4 is a computer simulation that represents learning as a social process of belief revision by a number of interacting agents. Each agent holds a set of beliefs about an aspect of the world that can be investigated empirically. The formation and revision of their beliefs is modeled as a process mediated both by making experiments and by communication between individuals. Agents are defined by attributes including: confidence in one or more of a range of hypotheses, variable sensitivity to the results of their own experiments and to the opinions of other observers, the ability to make decisions about whether to make experiments or to consult other agents and to make further decisions about which experiment(s) to perform or which agent(s) to consult. The paper describes assumptions underlying each of the basic features of the model, its application to the 4-card selection task of Johnson-Laird & Wason, an extension to this task to relate it more firmly to a real discovery process and some initial calibration results generated by the simulation of these experiments. The extension involves the restriction of communication and access to experiments. The results are compared with the related psychological experiments of Gruber. Further simulations including the comparison of the simulation’s behaviour for specific historical events are in progress.

Addis T.R, Townsend Addis J.J.

Systems Research & Information Science, Vol. 7, pp 263-291

ABSTRACT: There are two major reasons for the value of a functional schematic interpreter as a new programming language for constructing models. These reasons are the capacity of a functional language to express knowledge that has a formal representation on a computer and the rich informal semantics that can be attached to a set of diagrams (a schema). It is this informal semantics that help designers manage the complexity of model design. The combination of formal and informal semantics provides an elicitation, a specification and design language that can be run. The effect is the improved performance of system design through a path of minimal error. The design of complex models is shown to be further managed by two other strategies: the introduction of conceptual levels and the segmentation of design into functional domains. The notion of normalised objects is shown to be useful. However, the usefulness of this approach can be limited for some modelling problems. This limitation is illustrated through the design of a Roman Numeral calculator. Finally, the usefulness of a functional schematic interpreter as a specification and project management tool is also proposed.