Verifier and knowledge
A main theme in our research has been the Verifier approach for identifying error in expert reasoning about difficult problems. We later expanded this into the knowledge cycle. This work has involved integrating methods from four areas, namely elicitation, representation, error, and transfer.
Elicitation is about getting information out of people. From our Verifier viewpoint, this is essential for gaining an accurate understanding of what is is going on in a field. For people working in client requirements, market research, usability research, and related fields, this is a difficult problem, particularly when dealing with knowledge that people find difficult or impossible to put into words. Our work here involved developing a framework for choosing and using appropriate methods to elicit different types of knowledge.
Representation is about how to structure and present information. This includes ways of visualising how people categorise information, and ways of finding the most suitable way of visualising a particular set of knowledge. Our work in this area involves mapping different types of representation onto different types of knowledge, to guide appropriate choices. It also includes using visualisations systematically to reduce misunderstandings, mis-communications, and error.
Error includes active errors, where someone has made a clear mistake, and passive errors, where someone has failed to spot a solution or opportunity. We’re especially interested in passive errors, with particular regard to long-standing problems, and to clients who don’t know where to start looking for solutions to their problems. By using the most suitable elicitation methods to gather key information about a problem, combined with the most suitable representations for the core issues, we can significantly improve the chances of finding errors, and of preventing them.
Transfer includes training and learning, as well as communication. We’re particularly interested in craft skills – the everyday skills that are usually beneath the radar of training courses and the education system. Our work indicates that these skills are much more important than has previously been recognised, and that they’re as important in advanced theoretical professions as in manual work. Using appropriate elicitation methods gives us a better understanding of what someone knows, or thinks they know, about the topic in question. Using appropriate representations makes it easier for them to grasp the new knowledge correctly. We combine this with using appropriate methods to reduce the risk of errors and misunderstandings about that new knowledge.
Another topic that interests us and that has numerous overlaps with the four main areas of the knowledge cycle, is human desire. This involves investigating the underlying regularities in what people desire and what they fear. This work brings together concepts from fields as varied as humour, sport, film and horror. Some of our work here draws on models of aesthetic preferences based on information theory, which provides a more rigorous framework for understanding subjective elements of desire. This overlaps with our interest in types of knowledge that people find difficult or impossible to put into words.
Fields that particularly interest us
Our work spans a range of fields. This section gives a brief overview.
Archaeology: Developing methods for quantifying technological complexity and technological novelty, plus a way of identifying handedness in the archaeological record. The quantification methods provide a non-judgmental way of identifying the necessary preconditions for an innovation, including new inventions.
Design: Ways of applying user-centred approaches from software development to general product and process design, to reduce the risk of error and misunderstandings, and to make the product or process easier and more enjoyable to use. This includes methods for investigating user reactions to a product that they find difficult or impossible to describe in words.
Film and media: A broad range of ways to analyse film and media products and processes, including ways of visualising structures and themes within a film script via Search Visualiser.
Gender: Metalanguage and concepts for discussing gender and related topics with more rigour and with more validity.
Human factors: A range of work relating to design, to risk, to usability, to error, and to other human factors. Much of our work here relates to ways of handling knowledge that people find difficult or impossible to put into words. This includes methods for eliciting that knowledge, and methods for representing that knowledge, as well as methods for reducing misunderstandings and errors.
Medicine: Expert reasoning, and improving communication between patients and medical staff. Our work includes ways of asssessing whether Artificial Intelligence might be suitable for a particular area; ways of spotting where medical researchers might have made mistakes in their reasoning about a problem; and ways of reducing misunderstandings between medical staff and patients.
Requirements: Rigorous and flexible methods for gathering, clarifying and systematising client requirements. One of our main innovations was the development with Neil Maiden of the ACRE framework, which maps different types of knowledge onto appropriate methods for eliciting each type of knowledge, with particular reference to semi-tacit and tacit knowledge (i.e. knowledge that people find difficult or impossible to put into words).
Software development: Designing and developing novel, user-friendly software. An example is Search Visualiser, which shows keyword locations within a text in a way that lets the user rapidly make sense of very large numbers of texts, and of very large texts. Another example is the Quality Indication Visualiser, which shows indicators of high and low quality via annotation of text and of images of artefacts. This helps people learn which features are considered indicators of quality by e.g. experts, or customers, or people from other cultures.
Usability: Designing user-friendly products and processes. This includes using systematic methods for generating novel ideas via parallel processing and via serial processing, combined with methods for selecting the most promising methods, and developing them from concept to design.