Thursday 12 July 2012

Kybernetes Keywords: maps and miscellaneous

Together with colleagues Magnus Ramage and Chris Bissell, I have been appointed a co-editor of Kybernetes (the first issue under our names will be January 2013).

At the same time as us taking it on the journal is moving over to an online submission system. and one of the first tasks we've had to do is set up the initial list of Keywords. Emerald gave us a list from previous papers, but that list contained 910 words.

Our first reaction when we'd thought about Keywords was that they have little value these days, since you can do a full-text search instead. However, on reflection we realised that there are things we'd use Keywords for, things that a full-text search doesn't replace. For example, we thought about looking for reviewers. When you get a paper and need a reviewer, you need to find someone who knows about the field, which ideally would be identifiable by the Keywords. Of course, 'the field' can means lots of different things, it could be 'cybernetics' in a sense which might cover anything and everything published in Kybernetes, or it could be much more specific. It's about the level of abstraction, and I think the point is that the level of abstraction of Keywords should, in our opinion, be the level that you use in identifying a reviewer. That might sound a bit circular, but I don't think it is, It is about defining the level of abstraction appropriate for Keywords.

910 words is certainly too many. Either the words are too specific, too narrow to help in finding a reviewer, or else they come from too wide a field, taking in topics that are not appropriate for Kybernetes. The 910 word list was a bit of both. It was mainly too specific, but also - in a few cases - covering some topics that we don't think are really appropriate for Kybernetes.  (There were also quite a few cases of alternative spellings or synonyms, which clearly should be removed.)  We thought that we would want about 50-100 Keywords, but when we tried cutting down and editing the list, we came to 127 words (see below).

Apart from the practical issue of choosing reviewers, this list serves - or should serve - to identify the field of interest of Kybernetes. It is a map of Kybernetes.  It will evolve - we can add and remove terms as we go along - but it should help to understand what Kybernetes is for.

One of the most useful things for me about this blog has been the list of labels. At the moment there's about 140 labels (see in the left hand column).  As time has gone along, some of the labels I previously used have seemed less significant and others have become more important. Again, you don't need the labels because Blogger can easily do you a search of the full text, but that's not the point. The labels are the map and the map defines the territory (see what I've said about maps before by clicking on the map label). The labels are defining what this blog is about, and are defining what I think the territory of information is.

The map is not the territory, but without the map the territory is incomprehensible. The territory might as well not exist. Maybe everything is miscellaneous, but you can't comprehend 'everything', you can only comprehend the map.

Appendix: the initial map of Kybernetes

Here's the list of 127 terms we are starting with. I should put in the disclaimer at this stage that we generated it by cutting down the full list of 910, and we did it rather quickly so as not to delay the site going live. As I said above, it will evolve as we add and remove terms. It will become a more useful map, better representing the Kybernetes territory.

Action research, Adaptation, Aesthetics, Algorithms, Analogue, Analogy, Approximation theory, Architecture, Art, Artificial intelligence, Automata theory, Automation, Autopoiesis, Behaviour, Biocybernetics, Biology, Biometrics, Biophysics, Boolean algebra, Brain, Business systems, Causality, Chaos theory, Citizenship, Classification, Cognition, Communication, Communications technologies, Complexity, Consciousness, Control systems, Creativity, Critical thinking, Cross-cultural studies, Curricula, Cybernetics, Decision making, Design, Dynamics, Ecology, Economics, Education, Emergence, Energy technology, Environmental management, Epistemology, Ethics, Evolution, Experimental psychology, Feedback, Fuzzy logic, Gaia hypothesis, Game theory, Genetic algorithm, Globalization, Governance, Heuristics, History, Homeostasis, Human biology, Image processing, Information, Information systems, Information technology, Information theory, Intelligent agents, Internet, Knowledge management, Language, Learning, Linear programming, Management, Management information systems, Markov processes, Mathematical modelling, Measurement, Modelling, Monte Carlo simulation, Music, Nanotechnology, Narratives, Networking, Neural nets, Nonlinear systems, Numerical analysis, Open systems, Operational research, Optimization techniques, Organisational structures, Phenomenology, Philosophy, Physics, Politics, Probability theory, Problem solving, Project management, Protocols, Public policy, Quantum computing, Recursion, Relativism, Risk management, Robotics, Second-order cybernetics, Self-organization, Self-referential systems, Semantics, Sensors, Signal processing, Simulation, Social change, Social networks, Social systems, Sociocybernetics, Sociology, Statistical analysis, Sustainability, System dynamics, Systemic thinking, Systems theory, Tektology, Telecommunications, Topology, Turing test, Viable Systems, Wireless, World Wide Web


Tony Hirst said...

Hi David

A quick read of this got me wondering about whether a bit of text analysis and/or network analysis might help?

Trivially, just looking to see how keywords have been used to date might identify unpopular keywords, and charting their cumulative usage over time might show up different interest areas over time.

Graph-wise, if you have a list of keywords by article we could build a map of co-occurring keywords and then look for natural clusters/groupings/pairings, as well as giving an overview over the spheres of interest?

If you have titles and abstracts available, they could also be mined for possible keywords, or used to identify words particularly associated in the free text with given keywords etc?

Just a thought... if you have lists of titles, authors, abstracts and keywords in a machine redable form, I'd be happy to have a quick play to see if I can turn up anything of interest?

Martin Hawksey said...

There might be some R code you can reuse from some work Adam Cooper showed off at the JISC CETIS conference identifying emerging and dedining themes (code at end of post)

David Chapman said...

Thanks, this would be an interesting exercise. As discussed on twitter, I'll see if we can get the data in a suitable form.

Magnus Ramage said...

Thanks - an interesting set of reflections. A further reason in favour of keywords would be the classic finding in cognitive science that "recognition is faster than recall". This had a big influence in human-computer interface design with the rise of MacOS and Windows in the 80s. The idea is that choosing from a list of possible is generally faster than trying to remember terms and enter them. This makes a list of keywords (or labels) easier to navigate than full-text searching.

To take a different analogy, it's like the difference between browsing in a bookshop and searching at Amazon - it can be a a lot easier to find *something* interesting in a bookshop, especially if you don't know exactly what you're looking for.

So I'm glad we've got the keywords for Kybernetes!