|
Magpie: a person who collects or hoards things, esp. indiscriminately.
This course is entitled "Informatics Research Methodologies". Note the plural. This is because informatics has imported the methodologies of many other disciplines. It's not just that informatics is applied to just about every other discipline --- although, this is true. It's that informatics needs to use the methodologies of other disciplines in order to solve its own problems.
As evidence for this claim, I'll list below some of the disciplines whose methodology informatics has adopted; I'll outline what that methodology is, explain how and why informatics uses it; and I'll give links to one or more informatics papers using the methodology in this way. I'll start with the more commonly used methodologies and then continue to the more exotic end of the spectrum.
Some people consider Informatics to be a branch of engineering (and others consider it to be branch of science). However, in these notes we are concerned with how Informatics borrows the methodologies of (other?) branches of engineering. It does this especially when it builds hybrid systems in which software is embedded into some hardware. The three most obvious examples are: computer hardware, which uses electronic engineering; robotics, which uses both electronics and mechanical engineering; and computer modelling of physical artifacts, where engineering methods may be used to validate the faithfulness of the model against the artifact.
This last example illustrates a general phenomenon, which we will meet several times below. Whenever we build a computer model of some thing, and there is another discipline that studies things of this kind, we are likely to have to use the methodology of that other discipline in showing that the model is valid.
Mathematics is ubiquitous throughout theoretical computer science and in the many branches of informatics where formal reasoning is required. It can be recognised by its use of formal definitions, lemmas and theorems, and their proofs. It is typically used to infer properties of algorithms, such as their complexity, termination, correctness or completeness. Similarly, it can be used to show the properties of some formal framework, such as a deductive theory, classification or grammar. It is also used to show the inherent properties of some computational task, e.g., the NP-completeness of tautology checking. Used in a more informal way, it often crops up within other methodologies, e.g., engineering ones.
The use of statistics and probability is also widespread throughout Informatics. Statistics, for instance, is often used to analyse the results of experiments, e.g., to show that some observed effect is (in)significant. Probability is also widespread in theoretical models to represent uncertainty. For instance, it is the mathematical underpinning of neural nets and other forms of machine learning. It is used in knowledge representation to qualify the certainty with which facts are believed. It is also used in mathematical models of both computing and natural systems to represent the probabilities that certain events will happen, e.g., that some proteins will be produced by a cell in system biology.
In experimental psychology, experiments are performed on humans or other animals and the results analysed, often statistically. Informatics finds at least two main uses for this methodology. Firstly, in cognitive modelling, the faithfulness of the model has to be validated by comparing its performance to that of humans doing the same task. Secondly, in human-computer interaction, the usability of, say, an interface has to be measured while real users are using it. This methodology is also so important that we are devoting two guest lectures to it (to be given by Frank Keller) on experimental design and on cognitive modelling.
In philosophy, conceptual analysis is often used to clarify the nature of tricky concepts. It uses thought experiments to test the different possibilities, usually ending with a proposal. It is especially used in cognitive science and artificial intelligence to clarify concepts such as agency, time, belief, meaning, causality, obligation, etc.
Linguistics frequently use something similar to conceptual analysis to probe the structure of language, e.g., to propose grammars for a language. There thought experiments might consist, for instance, of both well-formed and ill-formed utterances and the circumstances in which they might be used. Computational linguists use this methodology in the building of natural language understanding and generation programs.
Sociological methodologies play a role similar to experimental psychology, but at the level of groups of humans rather than individuals. Two areas where this is useful are in understanding and improving the way in which teams of programmers develop ICT systems, and in understanding how new ICT systems might be or have been received by their intended users.
Biological methodologies also play a role similar to experimental psychology, but at the level of the wetware of the body and especially the brain. This includes computational modelling of aspects of the brain, which can be regarded as a branch of cognitive science, but it also includes attempts to understand how the brain works as a source of inspiration for computer processing, sometimes called "nature-inspired computing".
The medical methodologies case is essentially the same as the biological one, but usually restricted to computational modelling where the aim is to model some disease or its cure.
Business studies methodology is similar to sociology, but targeted at the case where the system is a commercial product and the users are customers.
Legal issues enter into many aspects of informatics, such as: the misuse of computers, e.g., hacking, viruses, etc; intellectual property rights; legal liability for computer failures; etc. These considerations obviously overlap with the sociological and business studies ones, and might form part of software engineering. The actual methodology is similar to conceptual analysis, where `thought experiments' are conducted and analysed in the context of the applicable laws --- and might be real legal cases.
The economics methodology case is similar to the biological one, where the object being modelled and validated against is an economic entity.
Historical methodology can play a similar role to sociological methodology when it comes to analysing the history of some ICT system(s), for instance, to discover for what reasons it succeeded and/or failed, and what lessons for the future might be drawn from this.
I put this one in to demonstrate the breadth of my claim, but don't expect art criticism to be a regular occurrence in papers you read. The role of art criticism in the example paper below is similar to the use of experimental psychology to assess usability in human-computer interaction. The product that users are to evaluate is the graphical depiction of art works, so the quality of the product is inextricably linked to the message the artist was trying to convey.
Alan Bundy|
Informatics Forum, 10 Crichton Street, Edinburgh, EH8 9AB, Scotland, UK
Tel: +44 131 651 5661, Fax: +44 131 651 1426, E-mail: school-office@inf.ed.ac.uk Please contact our webadmin with any comments or corrections. Logging and Cookies Unless explicitly stated otherwise, all material is copyright © The University of Edinburgh |
