Peer comments from JenD

[jendatx]

LOVE your choice of topic @emrapport! What an important dataset, with real potential impact.

My summary of the gist of your idea would be: Hidden insights in the raw text of police misconduct complaints could be exposed for analysis by means of data science, and these could help inform community actions to keep our family, friends and neighbors safer.

Here are my minor suggestions:

Slug

I'm not sure if the slug will ever be presented separately from the title, but (just in case), you might want to use "policing patterns" instead of just "patterns." Otherwise your slug might be indistinguishable from other slugs dealing with text patterns.

Keywords

I suppose you could consider also "police brutality", "police abuse", and "police misbehavior." There might be a way to determine what could help your project get found more frequently by using some of Google's SEO tools to see what most journalists look up, etc.

Middle

Nice explanation here.

Regarding text patterns

You could include a couple of examples, like for root word perhaps say that "strike" may need to be grouped with "struck" and "striking". Would a word like "hit" be grouped with words like "strike"? Would "cuss" be grouped with "curse?" I think one or two examples would drive home your point here.

Regarding relevant tags

Consider expanding this to a bulleted list so you could say "Locations: sidewalk, street, apartment, house, parking lot, park, business, etc." I don't know what's available in the data, but things like police rank (at time of complaint) with a descending list of org hierarchy, time of occurrence (24 hour clock, day of week, but maybe also shift), witnesses present (Public bystanders: 3; Cop bystanders: 1), could be in the bullets. This way, people who know the dataset (and the police force, and the community members) could easily scan the tags and point out any missing tags they'd like to see analyzed.

End

I always wonder if analyses like these would better reach their intended audiences if they ALSO included in their data sets incidences where police are injured/killed, alongside the data about residents injured/killed. That would help send the message that the data scientists seek answers to safety for both "black lives" and "blue lives." This might be especially important if the public will be given methods to explore the data on their own, as proposed here. It would anticipate push-back and dismissal from those who consider themselves "pro-law-and-order." And I think it would be more effective overall: if such a comprehensive analysis identified that there's an anomalous level of both complaints and violence (with residents and police hurt more on 3rd shifts in apartments, for instance), then solutions could be sought for these circumstances specifically. The aim is to keep all parties safe while justice is pursued. It could also help with bright spot analysis, for instance highlighting that graduates of XYZ police academy, even while sometimes taking minor injuries to themselves, have never had complaints against them; perhaps they received particularly effective de-escalation training that needs propagation through the wider force.

Thank you for letting me read this, @emrapport - I bet there's some really impactful insights to be had here! JenD

[emrapport]

Thanks, @jendatx. This is really useful. I wouldn't have thought of looking at google trends to see what other word combinations are searched for a lot - I did that and found that "police misconduct" is searched far less frequently than some similar word pairings. I liked your examples about how to make the middle more readable/specific and I'm implementing those as well. The idea about adding data about police injuries is interesting to me, as you're right, projects like these have the possibility to be very polarizing. I'm not sure if data on police injuries is ultimately the right approach, though, as these police misconduct reports include a wider range of scenarios than injuries/killings alone, so police injury data doesn't feel like the right counterpart. The correct counterpart seems like it would be civilian reports on positive interactions police, but I don't think there are reporting mechanisms for that, or if there are, I'm not sure if they are used. I'm definitely going to put more thought into that area. Thanks!

[emrapport]

Summary of issue: Changes made: updated slug, added more keywords, gave examples for text patterns, elaborated on relevant tags Suggestions I didn't use: adding data about police injuries/killings, for reasons stated above. Still considering ways to make project less antagonistic to police force, and provide useful context.