Increased airplane safety by IGNORING final outcomes October 1, 2007
Posted by Paul Duignan in : Accountability, Outcomes theory & the news, Outcomes models , trackback
The New York Times carries a story today, ‘Fatal Airplane Crashes Drop 65%‘. It reports that fatal airplane crashes in the US have dropped by 65%. The death rate in 1997 was one in 2 million whereas the death rate now is about one in 4.5 million. I think the level of airplane safety is one of the great administrative, regulatory and engineering achievements of our time. It shows what can be done when people are serious about managing negative externalities - bad things that happen in the course of selling goods or services. This success story illustrates an important, but seemingly counterintuitive, principle of outcomes theory - deciding whether or not a decision was a correct one often does not depend on the final outcomes from that decision. In other words, you can improve airline safety by ignoring final outcomes!
How has the aviation industry and its regulators actually achieved this improvement in its safety record? One of the examples of what has made a difference is given in the article - analysis of ‘black box’ flight recorder data from planes coming into airports has shown where there are unusual approach patterns. For instance at one airport, planes were consistently coming in too fast and at a steep angle - called unstablized approaches. Once this was identified, the Federal Aviation Administration changed the approach procedure and the airport introduced better guidance systems. From an outcomes theory point of view, the interesting thing about using this method is that as the article says ‘nearly all unstablized approaches end with a safe landing’. So in terms of a strict analysis of final outcomes, these flights were all totally successful. It’s only by ignoring the fact that they achieved their final outcomes and investigating things that ‘might have’ resulted in failing to achieve final outcomes that progress is being made on airplane safety. So decisions which appeared ‘right’ in terms of final outcomes - there was no crash - appear ‘wrong’ in the light of an analysis of the way in which the planes were being flown into the airport.
As the article says, ‘Analyzing data from safe flights is a reversal of the historic practice, which is to go out and ‘kick the tin’ after a plane crash, looking for clues.’ Putting a great deal of emphasis on ‘incidents’ rather than just disasters is now a standard approach throughout many types of safety administration. Incidents are of great interest to investigators because the provide a much larger sample from which to draw conclusions about how to improve safety that just looking at crashes.
So, if outcomes theory tells us to ignore final outcomes when it comes to deciding whether or not past decisions were right or wrong, where do final outcomes come into the picture? Information about final outcomes is used not to retrospectively judge whether or not a decision was correct (after all the decisions the pilots of the unstablized planes made generally ended in safe landings), but rather to improve the outcomes model that we use in the future to help us make future decisions rather than judge past decisions. So the reason the investigators in this case were focusing on unstabilized approaches was that a study found that such approaches were a factor in two-thirds of 76 accidents and serious incidents worldwide during landings from 1984 to 1997. This past data about outcomes informed the outcomes model and then that outcomes model has been used over the last 10 years to change the decisions which were made, this has led to less unstabilized approaches to US airports and it is believed, to less crashes.
The most radical application of this sort of approach is when it is applied to areas where people are legally held to account for their decision making. A good example of this is in the alcohol-impaired driving area. There, interestingly, punishment has been detached from final outcomes. An alcohol-impaired driver is punished not for the final outcome that they have killed or injured someone while driving, but because they have made a decision which increases the probability that they might kill or injure someone while driving. So their decision is regarded as wrong, and punished, regardless of the final outcome. The areas where this type of approach to legal sanctions are the most socially acceptable is where a violation of one of the lower steps in the relevant outcomes model - which is the basis on which they are being being punished is well validated. Areas which are under heavy monitoring and investigation and where the causal processes which lead to final outcomes are relatively transparent lend themselves to such approaches.
Paul Duignan (outcomesblog.org)
Comments»
no comments yet - be the first?