The Art, Craft & Science of Detective Work
Growing Meta #25
Hi all!
Growing Meta is a short weekly letter tackling meta angles, such as complexity, analysis or the future. Sometimes, it is visualized.
It goes out to 126 intellect seekers. Thank you for being here!
I mostly work with structured data (data you can look at in excel), which is a slightly easier job than working solely with unstructured data. The burden lies in finding the questions that really matter. Let’s say you’re tasked to design customized data packages that optimize for a telecom’s profitability, the question of “what local events incentivize users to consume more data?” is more impactful than “what is the effect of lunar cycles on data consumption?” — even though the latter is just as interesting.
You are lucky if you have a question. Sometime the problem is so complex and hidden that you would need to think laterally to connect the dots. In data science practice, we do explorative data analysis (EDA) for that. We do EDA systematically (or so we think) until we get an interesting hypothesis.
I always thought of a data scientist’s role as a detective. Detectives collect evidence and are forced to connect dots (i.e. ask the right questions). Finding the right questions/hypothesis is the same with detective work (criminal investigation). There is an active debate on whether detective work is an art, craft or a science [1]. In the media, detective work is usually illustrated as art, I presume it’s because it would be boring to watch Sherlock Holmes measure things by the book without any resourceful gimmicks. Even though the author believes detection should be science:
Detection is, or ought to be, an exact science, and should be treated in the same cold and unemotional manner. - Arthur Conan Doyle (the Sherlock Holmes series author)
When detective work is art, the detective follows their ‘instincts’. They act resourcefully in seeking out information (e.g. going undercover to investigate suspicious witnesses), and they think creatively about what type of information can support the investigation. Currently, the stage of data science practice is ripe with opportunities for artists. Data scientists who think laterally increase the opportunity for added ‘amazing factors’ to decision makers.
Detective work is seen as a craft when the detective utilizes their experience at solving cases in organizing the investigation process in a way that suits case at hand. Basically, a crafty detective is someone who has a structured mind and good priors (practical experience).
When detective work is science, there is a clear method for collection and analysis of clues. With increasingly advancing tools like offender profiling, forensic analysis, and information tech, guesstimating and assumptions will be a things of the past (we hope). When discussing this with my sister, she argued that we can’t ever rely on science alone because of game theory: a smart criminal will try to shape the scientific investigation. She’s right. The same thing happens in data science, we can’t rely on macro views of the data, and we certainly cannot objectify it. Most likely there will always be errors in your data (unintentional and intentional).
I turned to detective work for inspiration to become a better data science, but found that even detectives make grandiose mistakes due to lack of methodology and to inherent cognitive biases. The case with criminal investigations is that there is only one correct answer in a wide array of possibilities — starting with the wrong hypothesis can, and has, ruined lives.
Read this beautifully summarized horrifying story on a man whose wife died:
“In September 2008, close to the village of Loftahammar in Sweden, Ingemar Westlund (68) found the body of his wife Agneta (63) by a lake. He claimed he had last seen her alive a couple of hours before when she took the family dog out for a walk in the forest. He had just finished his weekly lawn mowing with his tractor. When she failed to return he went out to look for her. He immediately called the police and due to the nature of the injuries on her body a murder investigation commenced. Mr. Westlund told the police that he heard what seemed like a big splash from the lake just before he found her. Animal hairs which did not belong to her dog, were found on her jacket. They were not subjected to further analysis. Instead, based on a number of linear cuts containing fragments of grass on the deceased’s chest the police hypothesised that she was injured and asphyxiated by some mechanical cutting device such as a heavy lawnmower. Though no forensic evidence such as blood, hair or DNA, was found on Mr. Westlund’s lawn mower, and even though he had no motive or other irregularity in his statement, he was arrested on suspicion of murdering his wife. He was held in police custody for ten days and under formal suspicion for five months. In the interviews, he denied killing his wife and stuck to his initial statement. The police held several case reconstructions with the tractor lawnmower but never succeeded in getting on top of something that came close to the size of a human body. The case remained status quo for more than a year. At that time, however, the local gossip and unrest following the incident had made Mr. Westlund flee the village he had lived in all his life. A year and a half after the arrest, a forensic analyst at the National Forensic Centre found a clip on the Internet of a moose attacking a woman outside a shopping market in Alaska. He then engaged an elk specialist and together they found large amounts of elk saliva on the deceased’s jacket. The local police never informed the husband or the media about the elk theory. Later on, some prints from a pair of sample elk hoofs were paired with the injuries on the deceased. The comparison showed a near perfect match. Finally, at a press conference the case against Mr. Westlund was dropped. A local journalist later informed him about the decision. The police and prosecution service never apologized for what had happened and claimed they did nothing wrong. Mr. Westlund remarked: “My life is ruined. I have been dragged through a nightmare – a modern-day witch-hunt beyond my wildest imagination.”” - Source [2]
Until next week!
Najla
p.s. if you liked this, can you share Growing Meta with someone who likes Black Mirror?
[1] Tong, Stephen, and Ben Bowling. "Art, craft and science of detective work." The police journal 79.4 (2006): 323-329.
[2] Fahsing, Ivar A. "The making of an expert detective: Thinking and deciding in criminal investigations." (2016).