I've enjoyed using it for coming up with the structure of a project. I'll ask in search mode for structures of other similar projects if I'm not sure. I also enjoy making human-readable .md or .txt documentation files for myself very quickly with it.
I believe this is the case also. With a well enough performing AI/ML/probabilistic model where you can change the model's input parameters and get a highly accurate prediction basically instantly, we can test theories approximately and extremely fast rather than running completely new experiments, which will always come with it's own set of errors and problems.
I like the message, but I feel like this is bad data visualization. The width of each group of people is not the same, so it's somewhat meaningless to visually compare groups without being able to see the raw percentages. For example, the "Many Adverse Experiences" group is stretched to be longer than the other groups so that proportionally fewer people in that group appear to be a larger proportion than the same proportion would be in other groups because they're not as wide.
I'm torn. On the one hand, I agree with your remarks. On the other hand, I strongly appreciate the attention to detail in:
- Actually keeping individual datapoints all the time, clickable and with full details, and just moving them around to form different charts;
- Making the icons consistent with data - based on a few random instances I checked, the person's body shape and hairstyle correlated to biometric parameters in the data set.
I don't even think that the message is likeable. "Oh no they don't go to college!" is schoolmarmish and patronizing. "College is for everyone!" and "you're not really an adult until you're 25!" have done an awful lot of societal harm.
As a college non-graduate, I think that is leveraging the strong data that for most people a college degree is a huge net benefit is reasonable.
As someone who was once <25, I think that version of me is stupid in a wide variety of ways. I hear you that it can be negative to divide things that way, but it seems reasonable to say “after you are either a non college graduate with a number of years of experience or a college graduate with ~2 years of post-college experience.
I hear you, though, it’s hard to sort people into buckets.
> As a college non-graduate, I think that is leveraging the strong data that for most people a college degree is a huge net benefit is reasonable.
As another college non-graduate (although I'm currently going back to school). Have we ever figured out which way causation goes on this one? Does college actually have that much benefit, or do people who are motivated tend to go to college more?
As another college non-graduate that has friends that are graduates and non-graduates the graduates tend to live more fulfilled lives, regardless of income level.
People with support to go to college are more likely to go to college, less about motivation.
I think there needs to be more support for students failing/dropping out of college.
Also, the visualization doesn't update well when scrolling back and forth; and the grouping is bad -- "bullied" is listed as an adverse condition, but is also shown as a separate grouping; and the way it's displayed for "Seen someone shot with a gun" is backwards, implying that the vast majority have seen that. Too bad, because it otherwise seems like an interesting study.
Social sciences is not value-free. In reality the most important indicator of "at-risk" is previous involvement with social services and mental health professionals. Usually because these experiences tend to be so bad that the kids involved start to hide problems, or even attack anyone involved with social services. And THEN they get into a negative spiral. It is not the first time they get into a negative spiral, except now their experiences with mental help are so incredibly negative they fight to remain in the negative spiral, sometimes to the point of physical violence.
Likewise, these professionals hide that almost all experiences kids have with social services are negative for the kids. Now I suppose you could say the above is an example of that, but really, it goes further. Kids seek help with homework, and only get berated by someone that couldn't do the homework themselves ...
Studies keep pointing out that social services is exactly the wrong approach. What makes teachers, and social professionals good is excellent subject knowledge, combined with basic psychology. NOT the other way around. And in practice every mental help professional I've ever seen thinks they know what to do, and when pushed fail to produce even basic psychological facts, or outright deny them. I like to think you can explain this that when push comes to shove our minds are trying to solve problems in the real world.
The majority of mental problems are someone failing to solve real world problems, and repeatedly failing to influence the outcome. A little bit of psychology is needed to get them to try again ... and a LOT of knowledge of the real world is need to make sure the outcome is different.
I thought the visualization was awful. Prose and some (non-animated) charts would do a much better job, and suit scrolling/scanning back and forth much better.
I understand the motivation of trying to (literally) humanize the data points, but it would have been much more successful if there were vertical groupings as well as horizontal ones.
Right now it's 3 buckets + colors, but you could literally make it monochrome, make it an actual grid, then you could see which cells are completely empty, which is impactful.
I think you're reading it right.
They have the color key correct but the key for which side is seen vs not seen is incorrect.
It should be
<--Seen someone shot ... Not seen someone shot-->
Agreed. Spent a couple minutes trying to figure out how I was reading it wrong for several of the categories - sometimes it is correct, but often it is not.
I know that the author is trying to argue that minorities are at higher risk for bad outcomes, but it feels intellectually dishonest to use the same colors for white and rich, or black and poor. If white people can be poor and black people can be rich, you can't overload the color to reinforce your bias.
Plus, that whole section seemed to be sorted in an incoherent way.
Agreed, not least because:
- area-based visualizations make the effect hard to distinguish; bar charts or data clouds with numbers and confidence intervals would have been way more immediate.
- the colors make the negative group (usually) more visually prominent, since it has higher contrast with the background, exacerbating the area-estimation problem. (e.g. me wondering, "are there more overweight pink people as a fraction of pink people?")
Came here to say similar - making the page extremely wide helps a big by making the rows more similar but ideally consistent scale and number of rows should be maintained so we can see a column-to-column width comparison of the data points.
The visualization is a good iteration on trying to get complex papers distilled into a digestible format. That was nice.
I'm not super sure how I feel about the message though as it operates on a handful of really big presumptions. I'll share my own bias to save everyone the tldr on where I'm coming from: I'm a parent advocate. I think the nuclear family is the backbone to society and that much, if not every, societal ill can be linked to the destruction of the nuclear family. Parents matter, and I agree with the general conclusion that we need to focus TREMENDOUS effort into raising children in a loving and safe way. If you are still reading, consider also that I'm a 3rd generation son of Mexican immigrants. I grew up in a lower economic class background in Los Angeles county during the 90s. I grew up shoulder to shoulder with many of the people included in this study.
The first is that it's somehow a bad thing not to go to college. The trades by now are a known lucrative path with significant upward mobility, especially as we consider entrepreneurship. This is, in my experience, hand in hand with a lot of cultural practices that just doesn't get captured in these types of sociological studies. I can personally attest to the increased risk tolerance that a lot of cultures have towards starting a business or joining a labor based trade. Food trucks, car washes, detailing services, maid services, laundromats, dry cleaning businesses, convenience market franchises. In the privacy of your own head, and without fear of judgement from your HN peers, I invite you to honestly consider the ethnicity of the people who own these businesses. See my point? The mobility is there. These aren't "bad" lives. They're different. These people also have different standards of living. Most people who are immigrants or 2nd to 3rd generation of those immigrants don't want a multi-hundred thousand dollar life. Just speaking from personal experience here, most lower class migrants see the prospect of making that much money in America as foreign and unsafe. Maybe this furthers the point that not everyone should or can be a doctor/lawyer/FAANG-engineer.
The second presumption is that "abuse" or "adverse experiences" is able to be categorized by the researcher's definition. Again, we're dealing with people of different cultures who have different standards for living. We're overlaying our own "refined" terminology of what constitutes "abuse" or "danger" to them and drawing conclusions. Worse yet, we're saying that those same conclusions are correlated to the conditions that they experienced, regardless of how they themselves would classify it.
"High risk" is a highly contestable term, especially as the diversity of subjects increases. Maybe it's a good thing that mom divorced the man who was never around. Maybe mom was sleeping around and dad found out? Maybe mom remarried because dad died. Either way, non-intact households are being labelled "high risk" in a general sense.
"Being held back" as a bad thing is contestable. Some kids fall in that weird Nov-December enrollment period and make it through by being the oldest kid in their class. This isn't typically a good thing. The threat of being held back a grade is also encouraging for those who take their schooling seriously. Should it ever happen, its a serious kick in the pants for kids to wake up and take this seriously.
"Suspension", again any type of school based discipline, is seen as a adverse event. Suspension protects the children of the school, it notifies the parents of the suspended that there is a __real__ problem with your child, and provides a significant deterrent from bad behavior. It's wild to me that anyone would think of suspension as a noteworthy heuristic for adverse experiences.
Thanks to anyone who made it this far, even those that will disagree.
> The second presumption is that "abuse" or "adverse experiences" is able to be categorized by the researcher's definition. Again, we're dealing with people of different cultures who have different standards for living. We're overlaying our own "refined" terminology of what constitutes "abuse" or "danger" to them and drawing conclusions. Worse yet, we're saying that those same conclusions are correlated to the conditions that they experienced, regardless of how they themselves would classify it.
I think in this case, it seems they did pretty well. They're not lumping in "people failed to use their pronouns" into it, but things like gun violence, violent crime, and bullying. Some kids might be made of tougher material and shrug that off better, but even for them if that's not an adverse experience, I don't know what could be. It seems like the researchers are using an appropriately conservative definition.
> Maybe it's a good thing that mom divorced the man who was never around.
Yeh, but now we're confusing propaganda that was designed to encourage women to leave abusers for something of statistical significance about another matter entirely. If there are more men who would have made the kids' lives better than there are men so dangerous it's good they were separated from their children, then it doesn't matter that some are bad. The fact that the father has divorced and is out of the picture puts them at a higher risk of poor outcomes.
> The fact that the father has divorced and is out of the picture puts them at a higher risk of poor outcomes.
Hundred percent agree on this point. My concession was that it's not always beneficial that the parents stay joined nor is it deterministic that a single father or mother is strictly worse off than an intact family with an abusive/negligent/not present parent. Ideally none would divorce, but we can't factor for that.
> Did a parent or other adult in the household often or very often… Swear at you, insult you, put you down, or humiliate you? or Act in a way that made you afraid that you might be physically hurt?
seems pretty clear to me, regardless of if something is considered okay in one culture but not in another, the question is was the experience humiliating, not did X happen, where X could be considered not humiliating in one culture and not in another.
That's an excellent observation. When I wrote this I was looking for the questions/heuristics from the study that produced these statistics. I couldn't find much. Do you happen to have a link by any chance? I'm sure others would find it helpful as well.
I've found that for questions that aren't super complex on things that are extremely likely to be in its training dataset (such as public documentation of popular Python libraries) the error rate is very close to 0%. Even compared to GPT from a few months ago I've found the accuracy of responses has increased dramatically.
I still believe anyone using these tools on a day-to-day basis should have a sense of "trust but verify."
