Skip to main content

Human computing

The term 'Human Computing' has been introduced by SBR to describe a broad perspective on computing to which Empirical Modelling can be seen as making a significant contribution. The idea behind Human Computing is well-summarised in the following abstract for SBR's seminar:

Whatever the physical medium for computation, for example, beads, wheels, signals, transistors or neurons, our thinking and language about computers have always involved a rich interplay between the metaphors of the machine and of the brain (or mind). Since electronic computing has become ubiquitous and pervasive this interplay has intensified: the dominating influence of machine meanings of computing is being challenged by more human factors such as personal, and 'personalised', computing, ease-of-use, multimedia and the 'network culture'. The term 'human computing' refers to a broad approach seeking to give full value to the essential human role in computing. Pre-eminent in such a role is the semantics of computation including the identification of problems, processes and entities, as well as matters of interpretation, viewpoint, purpose and significance. The Empirical Modelling research group is making technical contributions of principles and methods to this broader approach. Starting from human perceptions and sense-making activities that are subjective and unreliable the goal is to use computer-based modelling to explore and establish the assumptions, mechanisms and meanings that can be used to build reliable and well-understood systems. This movement from the unreliable to the reliable can be considered in a variety of ways that will be described and illustrated by some of the work of current and recent research students. Focussing on the human role in computing is not merely complementary to the formal aspects, it opens up new ways of thinking about these aspects and new ways of using computers.

The themes behind SBR's seminar are amplified by several models. They primarily relate to two kinds of generic business application: model-building for decision support and model-building for business process re-engineering [BPR].

Decision support: The Restaurant Management, Temposcope and Elevator design models (~wmb/public/simulations/sisyphus) illustrate some of the significant issues in decision support. There are several EM papers authored by SBR, Rasmequan and others on decision support themes: the two most relevant are "Strategic decision-support systems: an experience-based approach", which discusses the Restaurant Management model in detail, and "A new paradigm for computer-based decision support", which appeared in a special issue of the Decision Support Systems journal in 2002. The latter paper explicitly contrasts an experience-based approach to decision support with the logicist framework for decision support that is based around three characteristic stages:

  • problem identification;
  • development of alternative solutions;
  • selection amongst alternatives.

The EM model for elevator design discussed in section 4.3 of the paper illustrates how the re-engineering of a traditional expert system carried out by a research team at Stanford University can be viewed as mapping from a logicist framework to a model based on EM-style agency and dependency. This section also includes a comparison between the rather closed problem of 'designing an elevator' (as in 'choosing components with the appropriate specification from a catalogue'), and the more open-ended task of timetabling project orals. A complementary discussion, in which an EM approach to timetabling is compared with traditional timetabling software, appears in our paper on the Temposcope.

BPR: The Warehouse management and Prism models were developed with BPR issues in mind. Both are quite complex distributed tkeden models.

The Warehouse model, developed by Y-C Chen, was motivated by Jacobson's object-oriented software development approach to BPR, and originated from modelling specific use cases in the context of the form-based business processes such as might feature in a warehouse prior to the introduction of automation. The warehouse model raises some interesting and controversial issues. For instance, to what extent is it appropriate to model 'existing practice' in this fashion when considering the introduction of computers? to what extent can the EM model be construed as modelling the essential abstract observables that are necessary to conduct and audit the management activities of the warehouse personnel, rather than as simply a model of a paper-form passing process?

The PRISM (Process Rework Interactive Situation Model), developed by Michael Evans, complements the AMORE (A Methodology for Object-oriented Re-Engineering) paper cited in connection with systems development in Week 8. In the critique of AMORE from an EM perspective, a key concern is that business processes it is appropriate to see business processes as emerging from semi-systematic practice, rather than being specified as abstract behaviours and imposed upon practice, Such a distinction between perspectives on how complex business processes should be developed is a high-level counterpart of that between the traditional software development and EM approaches to building complex programs. Evans's PRISM model demonstrates how processes can emerge bottom-up through the gradual automation of activities that are first carried out manually. Such a view of process development is particularly appropriate to the study of rework processes, which - by definition - involve developing systematic fixes for singular conditions that arise 'by accident' in an otherwise well-defined process. The PRISM model illustrates a particularly intimate interaction between human and automated activity in the spirit of SBR's Human Computing. It also illustrates a variation on the Experimental Paradox, whereby what is at first experienced as a genuine accident outside the scope of the current process, can - subject to empirical understanding - be reclassified as a routine (even if in some respects anomalous) occurrence within a new and more sophisticated process.

The elaboration of the Human Computing concept through practical studies is complementary to a philosophical agenda concened in broad terms with the relationship between the human and the machine. The coda to SBR's abstract frames what is seen by many as a controversial claim: "Focussing on the human role in computing is not merely complementary to the formal aspects, it opens up new ways of thinking about these aspects and new ways of using computers.". A key idea here is exalting the study of the human-computer relationship from the Cinderella status that HCI enjoys in the science of computing to an altogether more central and fundamental role. The 'new ways' to which SBR alludes above are conveniently summarised here with reference to studies of technology that have been largely initiated from other disciplines outside computer science, and with reference to EM's contribution to the logicist debate in AI.

One international interdisciplinary association of researchers (in which there are - perhaps surprisingly - few computer scientists) is the Cognitive Technology (CT) Society. CT is concerned with the intimacy and irreducibility of the relationship between technology and our cognitive and social life. The EM group has contributed papers to three of the four biennial CT conferences that have been held since 1995, and hosted the last conference in August 2001. Our two papers in the proceedings of CT2001 - "Interactive situation models for cognitive aspects of user-artefacts" and "The computer as instrument" illustrate why EM offers conceptual support to the more human-centred view of computation and computer-related technology that is of the essence in CT. Chris Roe's digital watch model was introduced in the first of these two papers as a simple case-study to illustrate how EM can model state without first declaring a boundary between the user (or the mind) and the artefact (or the machine. For instance: Roe's counterpart of the statechart is to be construed as his mental model of the digital watch; the elaboration of this model can be seen as reflecting learning and discovery about the watch; extensions of the model can embrace a part of the context for use of the watch, as when it is used to record the finishing times of two runners in a race.

The logicist debate in AI is centrally concerned with how and to what extent human intelligence is linked to reasoning and to the use of formal language. Much of the CS405 module has been implicitly concerned with issues that relate to this debate. We began by drawing attention to the fault-line that runs between theory and practice in modern computing. The dialogue between Bradley and James about the logical status of pure experience has featured prominently at many points in the module. Methodical approaches and formal specification techniques in engineering design, software development, business process modelling and decision support can be seen as attempts to rationalise human activities that EM would regard as of their essence empirical and ultimately beyond rationalisation. Similar controversies have been evident in social studies in connection with the scientific status of disciplines such as psychology and sociology. The real controversy here is not about what can and cannot be rationalised, but with the status of what cannot be rationalised, and the nature of the relationship between reason and experience. Two WMB papers that relate directly to these themes are "EM and the Foundations of AI" and "Liberating the Computer Arts".

"EM and the Foundations of AI" discusses the big issues in the foundations of AI, as set out by David Kirsh, from an EM perspective. A major objective of the paper is to explain how the transition from private experience to public knowledge is illuminated by the EM account of model-building for concurrent systems. Studying this paper should be helpful in relation to understanding the concept behind an EM approach to concurrent engineering and educational technology. The paper also seeks to justify EM as an appropriate semantic framework for language (rather than to accept language as an appropriate semantic framework for EM). An important observation in this connection is that fundamental words that relate to our direct experience admit (in particular) first-, second- and third-person interpretations. This has been explicitly discussed in relation to 'personal agency', 'attributed agency' and 'circumscribed agency' in the CS405 module, and applies also to terms such as 'real', 'time' and 'state'. "Liberating the Computer Arts" discusses the extent to which EM can deal with what is not rationalised, as epitomised by human activities that we would informally classify as art and craft rather than science and engineering. It links EM with the representation of three elements that are characteristic of our experience and that contribute centrally to the difficulties of rationalisation: situation, ignorance and nonsense.