« Back to main Programme

Why does so much IT fail? — and what to do about it

Harold Thimbleby

Swansea University

Your first reaction to this title might be that your iPhone and iPad are wonderful (and Google Glasses are even better…), so if only we brought everything up to date all would be well. Indeed, we may think IT is wonderful, but the reality includes a wide variety of failures: from the heartbleed bug, international programs to computerise healthcare, and an embarrassing array of preventable disasters.

We start with the high-profile failure of the German World War II Enigma, the design weaknesses of which undeniably helped lead to the war's conclusion — the German cryptographers, despite high motivation, failed to spot design flaws in the technology. In hindsight it is now easy to see the reasons, but it provides a fascinating object lesson.

The same symptoms present everywhere today in all IT systems. Interactive behaviour is very hard to understand.

As the examples in this talk come up to date, they get closer to our own stakeholding, and we get more resistant to the issues.

Medical devices with embedded IT suffer from ubiquitous preventable design problems, but they harm people. In a world where preventable death in hospitals is the third major killer (in the UK, killing about 90,000 per year — just behind heart disease and lung cancer — eclipsing, by a factor of 40, road fatalities), medical IT is making a worrying contribution. An ubiquitous but completely avoidable problem is bad number entry design, where user errors (which one expects and should design for) have unpredictable effects on the devices.

We need rigorous approaches to user interface design and evaluation for safety critical applications, packaged as tools so they are easy and reliable to use. We have built stochastic tools and tools on top of PVS (a theorem prover) to help identify problems that are otherwise overlooked. Improved design is easy once you decide to do it, and we have shown that significant risk reduction is possible. The bigger issue is to persuade the world.

About Harold Thimbleby

Prof Harold Thimbleby FIET CEng FRCPE FLSW HonFRSA HonFRCP is Professor of Computer Science at Swansea University, Wales. His passion is designing dependable computer systems to accommodate human error, particularly in healthcare. See www.chi-med.ac.uk (CHI+MED: Multidisciplinary Computer-Human Interaction for the design and safe use of interactive medical devices) for a really exciting major project he is on. He introduced the idea of combining informal and formal usability engineering requirements in the 1980s and co-authored, with Michael Harrison, Formal Methods in Human Computer Interaction (CUP, 1990). Harold has over 270 peer reviewed publications, and he’s been invited to lecture in 24 different countries. He wrote Press On (MIT Press, 2007), which won the American Publishers’ Association best book award in computer science. Harold is emeritus Gresham Professor of Geometry (a chair founded in 1597), and has been a Royal Society-Leverhulme Trust Senior Research Fellow and a Royal Society Wolfson Research Merit Award holder. He has held visiting positions in New Zealand, South Africa and Canada. He was awarded the British Computer Society Wilkes Medal. Harold’s web site is www.harold.thimbleby.net, which includes movies, details of papers, programs and more. Swansea is situated next to the UK’s first official Area of Outstanding Natural Beauty, and has beaches that have regularly won “best beach in Britain” awards. Do come and visit!

Sponsored by

AdaCore Capgemini Engineering

Supported by

Capgemini Engineering
Rapita Systems Ltd
SDC Systems
Wind River
Harmonic Software Systems