It's a great question and one I struggle with, because essentially you are asking, what is the difference between a crime analyst and a data scientist? A crime analyst in a COMPSTAT world, is generating reports and answering very narrow questions--there was a stick-up at a gas station last night, how many stickups like that have there been in the last three months and where were they? The data science role is way more expansive. You could imagine the data scientist linking to court records to see what the crim history was of people who were arrested for this kind of stickup in the past, who their co-offenders were, whether they have co-occuring disorders, and what treatment resources are available in the local community to serve those needs. And linking across geographies to understand regional experiences. And analyzing state and national data to see how the local trend fits the larger picture
Having been police and written about what police do and what they can't do, this is spot on. Police are reacti, not proactive. Responding to the fallout of structures and systems already in place. And the competition of philosophy in departments and between agencies for what works what the role should be. Let's moneyball this. How policing has changed under militarization from community policing (my era). As in education, society might not like the answers of engaging in/building systems that marginalize, starve, oppress people into policing traps. Police won't like the answers and way forward either. There is an unrealized reality that police don't make us safe. Moneyball the answer for what does.
There's a certain undertone of Newtonian causation emerging in here, not in the author's precepts, but in the uses to which the data's lessons will be put over time. We're dealing with complexity, not linear, sequential, cause and effect. The data appropriately generates statements about conditions and influences that bend the probabilities and don't dictate inevitable outcomes, but it seems to me incomplete. But is the data you have the data you need? If what you find is what you're looking might what you miss be what you need? Without some complementary means by which to continuously explore the rich narratives of the events nourished in the culture people will settle for data-driven answers that answer the wrong questions.
I agree with this. I think it highlights the importance of thinking about data scientists separately as developers of the data, and as scientists. I've been chewing on this post for a couple of years, but realized I was thinking too hard about the scientist part and not enough about the data part, which is where I think the wins are today.
In policing, when departments do a poor job and crime goes up or remains persistent, they get MORE money and resources. One of the few industries/professions where this is the case. The monopoly comment is dead on. Some of the stuff that the Newark Public Safety Collaborative and the Center for Policing Equity in St. Louis are doing with data-informed community engagement (and the democratization of crime data, place-based analytics, etc.) is promising and is one of the few efforts to make community groups, the business community, etc. CO-producers of public safety.
I am watching those experiments closely, as I suspect many folks in our biz are. There are other experiments under way, or soon to be underway, that continue to build on these co-production efforts. I wish they were easier to evaluate, but there is a lot of early evidence of success--certainly enough promise to keep building.
Yup. I'm thinking about running all my posts through GPT-4 and generating a four-sentence summary to post along with the essay. Sort of not kidding....
John Roman humorously acknowledges his tendency to meander at the start of his essays, promising a direct approach this time, marking the 20th anniversary of "Moneyball: The Art of Winning an Unfair Game" as a pivotal moment that introduced the idea of data-driven decision-making across various fields. He critiques the policing sector for its resistance to adopting Moneyball principles, attributing this to its monopoly status, which disincentivizes competitive improvement, yet argues that policing could greatly benefit from integrating data science to enhance decision-making and operational efficiency. Roman delves into the nuances of language and power, reflecting on his contemplation of a comma in a statement about policing, to illustrate how subtle shifts in language can significantly alter perceptions and arguments towards policing practices. Concluding, he advocates for the substantial incorporation of data scientists into policing, suggesting this approach not only as a means to address current recruitment challenges and operational inefficiencies but also as a potential pathway to more equitable and effective law enforcement.
Possibly naive question from a non-specialist: how would you distinguish this proposal from the whole CompStat trend?
It's a great question and one I struggle with, because essentially you are asking, what is the difference between a crime analyst and a data scientist? A crime analyst in a COMPSTAT world, is generating reports and answering very narrow questions--there was a stick-up at a gas station last night, how many stickups like that have there been in the last three months and where were they? The data science role is way more expansive. You could imagine the data scientist linking to court records to see what the crim history was of people who were arrested for this kind of stickup in the past, who their co-offenders were, whether they have co-occuring disorders, and what treatment resources are available in the local community to serve those needs. And linking across geographies to understand regional experiences. And analyzing state and national data to see how the local trend fits the larger picture
Her dress definitely waves.
Having been police and written about what police do and what they can't do, this is spot on. Police are reacti, not proactive. Responding to the fallout of structures and systems already in place. And the competition of philosophy in departments and between agencies for what works what the role should be. Let's moneyball this. How policing has changed under militarization from community policing (my era). As in education, society might not like the answers of engaging in/building systems that marginalize, starve, oppress people into policing traps. Police won't like the answers and way forward either. There is an unrealized reality that police don't make us safe. Moneyball the answer for what does.
Well, OK. But.....
There's a certain undertone of Newtonian causation emerging in here, not in the author's precepts, but in the uses to which the data's lessons will be put over time. We're dealing with complexity, not linear, sequential, cause and effect. The data appropriately generates statements about conditions and influences that bend the probabilities and don't dictate inevitable outcomes, but it seems to me incomplete. But is the data you have the data you need? If what you find is what you're looking might what you miss be what you need? Without some complementary means by which to continuously explore the rich narratives of the events nourished in the culture people will settle for data-driven answers that answer the wrong questions.
I agree with this. I think it highlights the importance of thinking about data scientists separately as developers of the data, and as scientists. I've been chewing on this post for a couple of years, but realized I was thinking too hard about the scientist part and not enough about the data part, which is where I think the wins are today.
In policing, when departments do a poor job and crime goes up or remains persistent, they get MORE money and resources. One of the few industries/professions where this is the case. The monopoly comment is dead on. Some of the stuff that the Newark Public Safety Collaborative and the Center for Policing Equity in St. Louis are doing with data-informed community engagement (and the democratization of crime data, place-based analytics, etc.) is promising and is one of the few efforts to make community groups, the business community, etc. CO-producers of public safety.
I am watching those experiments closely, as I suspect many folks in our biz are. There are other experiments under way, or soon to be underway, that continue to build on these co-production efforts. I wish they were easier to evaluate, but there is a lot of early evidence of success--certainly enough promise to keep building.
The old TL:DR!
Yup. I'm thinking about running all my posts through GPT-4 and generating a four-sentence summary to post along with the essay. Sort of not kidding....
Blech.
John Roman humorously acknowledges his tendency to meander at the start of his essays, promising a direct approach this time, marking the 20th anniversary of "Moneyball: The Art of Winning an Unfair Game" as a pivotal moment that introduced the idea of data-driven decision-making across various fields. He critiques the policing sector for its resistance to adopting Moneyball principles, attributing this to its monopoly status, which disincentivizes competitive improvement, yet argues that policing could greatly benefit from integrating data science to enhance decision-making and operational efficiency. Roman delves into the nuances of language and power, reflecting on his contemplation of a comma in a statement about policing, to illustrate how subtle shifts in language can significantly alter perceptions and arguments towards policing practices. Concluding, he advocates for the substantial incorporation of data scientists into policing, suggesting this approach not only as a means to address current recruitment challenges and operational inefficiencies but also as a potential pathway to more equitable and effective law enforcement.
Yours was better! Much.