Ross Anderson's 15 minute talk in particular is essential viewing.
Update: Comments from Prof Anderson's slides below:
‘The Software Crisis’
· Since the 1960s about 30% of large private sector IT projects have failed
· We now have much better tools for managing the complexity but we haven’t cut the failure rate!
· We just build bigger better disasters
· So the failure rate is a function of corporate IT capability, appetite for risk and so on
The Public Sector is different
- According to Joe Harley CIO at dept for work and pensions public sector IT projects have a 30% success rate
- Public sector IT is £14 billion
The economics of dependability
- Rapidly growing field in which we study the effect of incentives on complex systems
- Most work so far is on individuals and firms acting as rational principals not on agency
Public choice
- Analysis of agency effects in public sector organisations – Buchanan, Tullock etc
- Insight – a modern state isn’t a personal kingdom any more. Ministers and officials incentives differ from those of Henry VII
- Ministers want to get in the papers and get re-elected. Officials want to maximise control
- Can this give us any insight into systems failure?
Private sector projects
- Suppose you are developing a new branch automation system for a high street bank. You need the directors to – commit to a specification closing down options early, shut up and keep out of your way for two years while you code up and test it, take hard decisions if need be when you deploy it and parts of it don’t fit how people really work
- How feasible is this for a minister
Ministers and projects don’t mix
- Ministers are not hired to do stuff but to resolve conflicting interests in society
- Most supplicants are rent seekers, most alarms are false alarms and will go away
- So a sensible minister will respond to most demands by flannelling and doing nothing
- Absolutely the last thing a minister wants to do is take lots of hard decisions up front and be explicit about them
Ministers and projects don’t mix (2)
- The second thing s a good project client has to do is to shut up for 2 years while the system is being built
- How practical is it for a minister to keep out of the limelight for two years
- And the third thing to accept quickly and publicly that some promises won’t be delivered, and change the business accordingly
Other toxins
- Power struggles, as with dept of health officials fighting doctors for control of patient health records (a fight that began in the early part of the 20th century)
- ‘Electoral’ (look good with the voter) projects such as ID cards
- ‘Security’ as with the Cabinet Office framework for information assurance
- All quite comprehensible to an economist – security economics is now well studied
The Result
- Public sector projects are started with eye-catching but unrealistic objectives
- The stated objectives get changed many times as ministers bid for attention
- Suppliers know this – they low-ball the contract and then make their money on the changes
- Normal public sector risk aversion doesn’t work. Ministers are in a tournament for promotion; chickens usually roost on the next guy’s watch; and suppliers know failures will be covered up
Public sector projects that work
- Congestion charge: extremely simple stable specification – ‘Every car that comes into
- Land registry systems – done by officials incrementally without any ministerial attention
One future
- Large complex systems nowadays evolve rather than being built from scratch (Office, Google, Facebook…)
- If government want to build ever more complex systems that actually work, it will have to re-invent itself around evolutionary systems development
Another possible future
- Economists realised in the 1960s and most politicians by 1989 that macroeconomics places limits on how much the state can do
- I believe that 70% systems failure rate is also telling us that microeconomics also places limits
- Societies are complex systems too and if you try to manage stuff centrally you can only get so far
- Maybe this is the next big divide between liberals and big government folk (left and right)
No comments:
Post a Comment