Hello everybody. My name is Hannes Spichiger and I’m presenting a short paper called The Use of Automated Systems for Rapid Decisions that I wrote together with my colleague, Timothy Bolle. We both are PhD students at the university of Lausanne in Switzerland.
So what are we talking about here? Nowadays automated systems are often used for classifying tasks so we launch automated systems on some sort of data to sort through it for us to do some sort of analysis grouping together elements within that data. Now ideally as those software systems are not always perfect, the outputs of those systems are reviewed after by an expert in the field. This is, however, not always possible if there is a temporal restraint.
So one scenario we can look at where such a temporal restraint is indeed in place is in an airport security context. The aim is to identify potentially dangerous persons that are attempting to take a flight. We can think of terrorists or persons related to terroristic activities.
So the technique that is proposed in this scenario is that we do a target extraction and an automated analysis of the suspect’s phone which will then give us an indication whether or not suspicious material was found on it.
We are not going to take into account legal and ethical considerations, are we, as we are more interested, really in how such a system works within a context like that. This is a brief schema of what we think about, the result of our analysis is then integrated into a decision process. Sadly, we don’t really have time to go into that in detail.
So what factors will have an influence on our final decision, whether or not we are going to keep this person for further investigation? Well, first of all, we have unevenly distributed priorities. Not every person has the same risk to be indeed a terrorist. This will depend on the flight’s origin and destination, as well as on the person herself.
Then there’s a number of system characteristics that will influence that the performance of the system, how well the system is understood by the person handling the situation, as well as the context of the traits that we are looking at.
Finally, there’s a perceived risk together with organizational and personal factors that will influence how the decision’s made, and all those factors are susceptible to change over time.
As a conclusion, a decision does not only just depend on the automated system itself, but also its surroundings, how it is implemented, how it is indeed used and understood by those that are in charge over the final decision, how much expertise they have and how well this expertise is applied to the current situation.
And finally organizations that aim to set up such systems should address all of those considerations, otherwise the system is likely to not work as well as intended, or even cause damage.
Thank you very much for your attention. If this presentation was of interest to you, we’d invite you strongly to read our short paper or to contact us by email. Thank you very much.