Formal methods and (ubiquitous) UI

Today and yesterday, we hosted a 'mini-conference' on HCI and formal methods. Harold Thimbleby and five (of his) PhD students were visiting, who all gave talks on what they were (planning on) doing. On the Newcastle side, three PhD students who Michael and I are involved in supervising, also presented their work.

Andy Gimblett (Swansea) talked about finite automata models for (discrete) UIs, taking an interesting (generative) approach on discovering sequences of actions that are part of a user's mental model of a device (i.e. it's UI). Jose Luis Cardosa Silva (Newcastle/Portugal) presented his work on Virtual Environments, comparing different modelling approaches for this kind of setting. Jennifer Pearson (Swansea) talked about her work on improving reading tools for electronic devices, i.e. by replicating some of the physical tools we use when reading 'real' books (e.g. bookmarks, annotations). She mentioned being interesting in exploring a more physical ways of interacting with digital content, which reminded me of [cite] and our initial forays into using a camera/projector-based lamp for similar purposes.

Chitra Acharya (Swansea) then presented her initial analysis of a modern hospital bed. These beds are really sophisticated devices, and the particular one she had analysed there were actually three separate physical interfaces (for the nurse, the patient and the attendant). Patrick Oladimeji (Swansea) discussed some ideas regarding the discovery of interaction errors in devices based on a simulation of a device UI he built in ActionScript. Robin Green's talk (Swansea) was about formal modelling for reactive programming. Abdulrahman Alhaziool (Newcastle) presented his first steps towards creating a generic approach to using/running formal models in the context of instrumented environments. Due to the post-grad congregation, I missed Haryani's talk (Newcastle) on Persuasive Technologies and Security.

All in all, it was a mixed but complimentary set of topics, and it will be interesting to see what will come out of them in a year's time. In the discussion, we talked about whether there might be a gap between "low-level" models for interacting with an individual device (such as a drug pump) and "high-level" models that deal with the overall workflow (such as how nurses attend to different emergencies). Some errors/issues might not be uncovered by looking at either one, for example, when two devices are used sequentially (or in parallel) during a particular workflow, there might be some interference that could result in incorrect dosages of certain drugs being administered to a patient.

Harold Thimbleby also gave a talk on understanding designed-in errors in devices (in the context of the School's research seminar). He was particularly focussing on devices used in hospitals (such as drug pumps and monitors), showing how even with the very simple functionality and interfaces they provide, there are still significant numbers of built-in errors that can be fatal. He was presenting a number of examples and some statistics to illustrate that large numbers of people are killed each year (at least partially) due to these relatively simple HCI problems, which can be detected (semi-)automatically.