It has been 10 years since the release of the film Minority Report. It depicts a future where police can predict (and, thus, prevent) crime. When I went grocery shopping recently, I saw the DVD on sale and couldn’t resist getting a copy.
In my mind, this film is very much related to my own PhD. It was released shortly before I started my doctorate programme at the LSE. Often, when I explained what I was researching (namely, the development of customer profiles to identify instances of money laundering in retail banking), I would be greeted with a ‘Oh… that’s just like Minority Report, then’.
Revisiting Minority Report after all these years made me think about customer profiling now vs. the movie’s vision of the future.
What are the similarities?
1. The goal
Both ‘Precrime’, the squad headed by John Anderton (Tom Cruise), and real-life predictive customer profiling rely on the principle that prevention is better than investigation. Being able to anticipate where and when crime might take place, means that actions can be taken to prevent it. These can be short-term actions such as stopping a terrorist just before s/he assembles a bomb or preventing an identify thief from accessing information or assets. It can also be long term actions, such as supporting the social and economic integration of young people at risk of turning to crime.
2. The problems
The social benefits of preventing crime are obvious in the film. The same happens in reality. I certainly feel more reassured to read that crime rates have fallen, than to hear that criminals have been arrested quickly after an incident. It’s the individual cost of false positives that is the problem.
In the movie, the ‘system’ predicts that John Anderton will commit a crime in the next 36 hours. A nightmare scenario ensues, with Anderton trying to prove his innocence and avoid being captured. In real life, too, many are temporarily unable to use their credit cards, are prevented from travelling, are stopped and searched, are reported to police or even taken into custody as a result of faulty predictions (i.e., the ‘false positives’).
False positives – in real life and in the film – occur because data can tell more than one story. Data is just data. It is a series of dots. We are the ones who see patterns in the data and interpret the correlations. Often, we see patterns in the data that simply are not there, as discussed in this great post by Tim Harford. And, as I discussed here, correlations are not causation.
What are the differences?
1. The process
According to the movie, in 2054 AD profiling will rely on a team of 3 psychic ‘beings’, the precogs. In real life, profiling uses automated systems, data mining, predictive analytics and, importantly, human intuition.
2. The future
When confronted with the limitations and the risks of the system, the authorities decide to abandon the Precrime programme, reverting to a focus on investigation.
Is that the destiny for customer profiling initiatives?
I don’t think so. I believe that predictive profiling is here to stay. It is too valuable a tool to be discarded, despite its flaws. Though, privacy concerns and legal challenges may change when and what data is collected, and how it is used.
Moreover, I think that, in the age of big data, we need more than computing power and statistic skills. We need common sense to explain the behaviours behind the data. And we need sensibility to consider the implications of the predictions.
Remember that story about Target sending coupons for baby clothes to a teenager? It was hailed as a sign that Target’s profiling systems are so good that they spotted that the teenager was pregnant before her family knew. Well… maybe the algorithm was good. But the marketing management was poor – it lacked the sensibility to think through the consequences of sending the vouchers to the teenager.
Do you agree?