Skynet for school shooting prediction

This sounds likely to be fraught with false positives.

In particular, the language a student uses during an interview can help distinguish a high-risk teenager [shooter] from a low-risk one, according to previous research Barzman directed. That study concluded that the former was more likely to express negative feelings about himself and about the acts of others. He also was more likely to talk about violent acts involving himself and violent video games or movies.

My clique in high school were the metalheads, barely passing nerds, and social rejects. Self-deprecation was the basis of our humor. Violent video games or movies was the basis of our media consumption. The only one of us every accused of fighting was a case of mistaken identity as the first and last (not middle) names matched the very common actual guy of another race.

Artificial intelligence is the tool of choice for this kind of stuff. I hope the research is light years ahead of where this article describes it.

Why does Microsoft hate Edge?

I have the default browser set to Chrome in the Windows OS. So, when I click a link in Outlook, it opens in Chrome. So, any time an application opens Internet Explorer, it is obviously something weird.

Okay, confession time, it makes me wonder if I am computer illiterate every time IE opens.

Since Edge will be three years old this coming July, I kind of expected the switch by now. Why are Microsoft applications opening links in anything other than the default browser? And why are they picking IE instead of Edge?

I thought I would fix the issue by uninstalling IE. In Control Panel > Programs > Turn features on and off, one can choose whether IE is installed. Unfortunately, that seems to have broken the ability of programs to use the default browser. Reinstalling IE fixed that.

ISP Maybe Not Neutral

It looks like maybe my ISP is forging their Ookla speed test results. Whenever I use that speed test, I get the speed I paid for, but performance otherwise still seems slow. Whenever I use other speed tests, I get results that are a quarter to a twentieth as fast.

The tests were not concurrently run. They were sequentially back and forth several times for consistent results. I’d need a lot more data to say one way or the other. I have not gotten to the point of running a thousand automated tests on every tester I can find.

Both consistently slow or intermittently slow Internet speeds might explain the issues streaming Hulu and Netflix. If it was intermittent, then in testing the speed the service might decide high speed means give me HD quality, but then a slow down causes it to start buffering and before it downgrades the fast speed is back.

How many records are there?

On the surface, it may seem like an easy question. Data Owners typically think this has an easy answer.

If their data followed a very simplistic model, then it would have an easy answer. Just “select count(*) from table;” and report the value.

Unfortunately for Data Custodians, the data is often organized in relational data structures. In my case, we almost always buy a product and are not provided or not allowed to see schema documentation. So, this question requires making judgment calls.

  1. What kinds of data do I have? Hopefully, the tables have decent names that are meaningful. Learn how to use the product and the basic concepts of what things are will come with it. Assuming the tables are meaningful, then a review of the table names will suggest where the likely important data is stored.
  2. Research elsewhere if necessary. People blog about what they know. They ask questions in forums.
  3. Experiment. Change the value in the application and look to see whether it shows in the table.
  4. Count. Get the number of records for the relevant tables.

 

Likelihood to read a tweetstorm

I’ve noticed something weird about my reading habits. I think only read maybe 1 on 25 tweetstorms authored by someone I follow. I will read about 1 in 5 retweeted by someone I follow.

Tweets appear with most recent at the top and oldest at the bottom. So, when I encounter a tweetstorm in my feed, the natural inclination is to scroll to the end and read them in reverse order. It is an unnatural threading.

When someone retweets a tweetstorm, it is just the one tweet with a notation that there is more in the thread. So, I click the tweet, it opens a page with the tweets in order where I can read them in order. It feels natural.

I guess I could click the last tweet from those I follow, get taken to the end and scroll back up to the top. That would put things in the right order.

Crappy Geofencing

A friend claimed geofencing is the solution to the problem of minor league baseball players clocking in and out for work. (Sure, I had other issues with the plan, but this is specifically about how if all that other stuff is wonderful, how geofencing would fail.)

The concept is that when one arrives at a destination, then a phone undertakes certain behaviors. Across a few phones over the past 4 years, I have been very, very underwhelmed.

  • Hit or miss: Location services are often inaccurate. Not like the wrong city inaccurate, but they often indicate I am at the wrong address which is just enough the action does not trigger. For example, a reminder app will trigger when I visit the grocery store to get things on the list. Except, it doesn’t half the time because it thinks I am at a business next door.
  • Battery drain: They usually demand one have GPS enabled. They also demand that you not be in a battery saving mode. So, you better have a great device with a huge battery or frequent access to recharging it.

I will admit this generally works much better today than it did 4 years ago. So, there is hope for the future. And the optimistic view is it probably just a couple years away, so the pessimistic view is more like a decade, so splitting the difference: 6 years.

Lying to Big Data

The reaction people have towards social media companies is to lie. This amuses me because self-reporting is well known as the worst data. The data scientists expect people to lie.[1] Which is why they ignore what you say about yourself and focus on your behavior.

So, you need to start having intentionally deceptive behavior. The problem is: if people like you all deceive in the same patterns, then the data points to the same place anyway. You have to deceive in novel ways others like you would never think of doing.

Good luck with that.

1. Christian Rudder’s Dataclysm: Who We Are (When We Think No One’s Looking) is about how as a data scientist for OKCupid, a dating website, he cannot depend on the honesty of people. He has data on what they say and compares it to what they do.

Fake News

There is an article floating around with a title, “Fake news sharing in US is a rightwing thing, says study.” I have not clicked on it because it sounds like fake news. The Guardian typically has clickbait titles.

Also, I am annoyed because if I wanted to do a study on how susceptible liberals are to fake news, then putting out a fake study like this is exactly how I would gauge how much they share because it would go super viral.