Using pattern recognition to avoid bad people

[drumbo]

Sometimes on the subway we see people behaving in a rude manner, e.g. blaring music from their phone without headphones. When this happens I take a look at the offender and take note of all of their qualities, including their haircut, facial features, and clothing. My brain then associates all of those qualities with bad people so that I can avoid them. Is this a reliable way of avoiding bad people, or are the correlations drawn invalid?

[iNow]

Humans find patterns everywhere, even when there’s no legitimate pattern there. Seeing a face on the moon or a connection between eating bananas and smelly socks.

You’re right to question whether your observations are valid. They’re probably not. You’re probably missing deeper and more important patterns, like the number of times an officer stopped and frisked them, or how many times fine people like you forgot to notice the good in them. 

Humans are also a bit random. You could notice the exact same pattern 10 times in 10 different people and get 10 different results. 

Just try always to find the humanity in others. Sometimes you’ll be let down, but rather often you’ll be pleasantly surprised. We kinda see what we’re looking for, so try looking for the good. 

[dimreepr]

4 hours ago, iNow said:

Just try always to find the humanity in others. Sometimes you’ll be let down, but rather often you’ll be pleasantly surprised. We kinda see what we’re looking for, so try looking for the good. 

Indeed, and so much more rewarding than living in perpetual fear.

[swansont]

Finding out if those correlations are causal would be science. A point you have stopped short of.

Without the science you just have bias and prejudice.

[Area54]

9 hours ago, drumbo said:

 My brain then associates all of those qualities with bad people so that I can avoid them. Is this a reliable way of avoiding bad people, or are the correlations drawn invalid?

There are no bad people; only bad actions.

(Moreover "bad" is a subjective judgement. Stalin probably viewed many of his actions in a positive light.)

[John Cuthber]

22 minutes ago, swansont said:

Finding out if those correlations are causal would be science. A point you have stopped short of.

Without the science you just have bias and prejudice.

As far as I can tell, he hasn't even got as far as correlation yet.
He doesn't seem to have collected data on "nice people"- whatever that may mean.

So he might, for example notice that "bad people" often have beards.OK, for a start noticing it might be confirmation bias.
It might be a bias in favour of women; perhaps he doesn't feel threatened by them so he always sees them as "nice" (there may be other reasons...).
And since women seldom have beards, he's going to form the conclusion that bad people have beards.

He's still going to come to that conclusion, even if all the angelically wonderful people in the world have bears. 
And, even then there's the fact that "bad people have beards" is not the same as "bearded people are bad"

Essentially the only thing he has got right is questioning his own judgement.
 

[dimreepr]

32 minutes ago, John Cuthber said:

Essentially the only thing he has got right is questioning his own judgement.

Baby steps...

[swansont]

1 hour ago, John Cuthber said:

He's still going to come to that conclusion, even if all the angelically wonderful people in the world have bears. 

That category will likely suffer from small-number statistics 

[drumbo]

2 hours ago, swansont said:

Finding out if those correlations are causal

Unnecessary. It is sufficient to establish that a statistically significant correlation exists in order to make valid predictions. For example, I can notice that crime rates are higher in neighborhood A than neighborhood B, and even if no causal relationship between crime rates and the neighborhoods can be established, I can still lower my risk level by simply avoiding that neighborhood.

[dimreepr]

7 minutes ago, drumbo said:

Unnecessary. It is sufficient to establish that a statistically significant correlation exists in order to make valid predictions. For example, I can notice that crime rates are higher in neighborhood A than neighborhood B, and even if no causal relationship between crime rates and the neighborhoods can be established, I can still lower my risk level by simply avoiding that neighborhood.

You're welcome to your fear, that seems punishment enough... 🤒

[swansont]

56 minutes ago, drumbo said:

Unnecessary. It is sufficient to establish that a statistically significant correlation exists in order to make valid predictions. For example, I can notice that crime rates are higher in neighborhood A than neighborhood B, and even if no causal relationship between crime rates and the neighborhoods can be established, I can still lower my risk level by simply avoiding that neighborhood.

Until something happens to change the actual cause, which renders the correlation moot.

If we use the beard example, since beards don’t actually cause bad behavior (or have a common cause with it), when styles change and people shave (or more people grow facial hair) the correlation will evaporate. But you’ll still be avoiding bearded people.

Science allows you to predict. Correlation requires you to collect statistics and is only apparent after-the-fact.

[drumbo]

2 hours ago, swansont said:

Until something happens to change the actual cause, which renders the correlation moot.

I agree with this, but from a game theoretic perspective it is in our best interest then to assume that the correlation has merit. If the correlation has no predictive value then I have discriminated unjustly, but there is little harm to myself, however it does then I have potentially avoided harm.

[joigus]

Inferring correlations when you see guys in the subway is no robust way to make serious statistics. It is well known that humans tend to apply filters and bias the results when they read too much into their everyday observations. For example, when people notice some guy who's dressing in a way they already disapprove, they tend to pay more attention when he's also acting in a way they disapprove. So you must be careful with such observations.

Edited September 17, 2020 by joigus

[CharonY]

The first requirement is a robust and unbiased means of data collection. Just observing and noting features inevitably results in a biased data set. Say, I for some reason hate country, I am more likely to find folks blaring country more annoying, for example.I might also in my mind memorize features I associate with country music listeners more strongly than other features. The result is that I then have, in my mind, evidence that country-related features are clearly associated with rudeness.

It is in fact very difficult to develop an unbiased method for recording observations even before one can do any analyses on the data set. Garbage in garbage out, so to say.

[Phi for All]

1 hour ago, CharonY said:

The first requirement is a robust and unbiased means of data collection. Just observing and noting features inevitably results in a biased data set. Say, I for some reason hate country, I am more likely to find folks blaring country more annoying, for example.I might also in my mind memorize features I associate with country music listeners more strongly than other features. The result is that I then have, in my mind, evidence that country-related features are clearly associated with rudeness.

It is in fact very difficult to develop an unbiased method for recording observations even before one can do any analyses on the data set. Garbage in garbage out, so to say.

As part of a more rigorous approach, I recommend noting the difference between country music, western music, and country-western music. Country is more hearth and home (Johnny Cash), western is about the Old West (Marty Robbins), so people who like those aren't really the problem.

Country western is crying in your beer over the wife leaving you because of the dent you put in the truck when you ran over your dog. None of the people who like country western have ever ridden a horse, but they all wear pointy boots and a cowboy hat, which should make pattern recognition easier. Now if we could only figure out how to identify the bagpipe players....

[John Cuthber]

13 hours ago, Phi for All said:

Now if we could only figure out how to identify the bagpipe players....

They play bagpipes.

[Phi for All]

2 hours ago, John Cuthber said:

They play bagpipes.

If you can't avoid that event horizon, it's too late for you. The law should require them to wear kilts.

[joigus]

Bagpipe players in the subway have always bothered me. But bagpipe players in the subway doing a headstand are even worse.

[Arete]

From a statistical perspective, your personal observations are anecdotes. That is not to say they are "wrong", but they are biased by your own personal predilections, where you personally travel, your habits, etc and so on. That is to say, that the observations of one individual, be it you or anyone else do not generate a representative sample. 

Non-representative sampling is statistically worthless. Further, purely correlative sampling is of limited worth - for a famous example, the increase in global temperature correlates with the decrease in the prevalence of pirates. 

I would say that using anecdotes about physical appearance to predict people's behavior would lead to a high degree of false positive and false negative results -and also a great way to amplify one's predilection for bigotry.

[John Cuthber]

6 hours ago, Phi for All said:

If you can't avoid that event horizon, it's too late for you. The law should require them to wear kilts.

This one has warning lights

 

[MigL]

Looks like he might have set his hair on fire a couple of times, and had to trim it back.

Besides, if you wanna hear some AC/DC pipe music you should listen to 'It's a Long Way to the Top ( if you wanna rock and roll )'.
'Thunderstruck' is stripper music ( so I've been told ).
 

[John Cuthber]

14 hours ago, MigL said:

Thunderstruck' is stripper music ( so I've been told ).

I hope nobody tells that to the busker.

[MSC]

Using pattern recognition to identify and treat sick people, is probably a better way to phrase this.

Bad is quite the strangest term to apply to an individual. I blame whatever maladaptive family and support culture they came from. Genetics has it's role in determining certain aspects of disposition in life, most individuals of most dispositions, given intervention and guidance can find moral functionalism somewhere in this wide world and wellbeing and security along with that. I don't care what anyone says, I want my surgeons to be at least a little bit sociopathic. Their hands don't shake. Same with dentists, but be punctual, if you are late they can be a little sadistic with even a simple cleaning. Learned that one the hard way!

 

On 9/18/2020 at 9:32 PM, MigL said:

Looks like he might have set his hair on fire a couple of times, and had to trim it back.

Besides, if you wanna hear some AC/DC pipe music you should listen to 'It's a Long Way to the Top ( if you wanna rock and roll )'.
'Thunderstruck' is stripper music ( so I've been told ).
 

Damn.... I must be a male stripper then... I could probably make it work. I fire spin! 

[J.C.MacSwell]

On 9/18/2020 at 11:32 PM, MigL said:

Looks like he might have set his hair on fire a couple of times, and had to trim it back.

Besides, if you wanna hear some AC/DC pipe music you should listen to 'It's a Long Way to the Top ( if you wanna rock and roll )'.
'Thunderstruck' is stripper music ( so I've been told ).
 

Something no true Scotsman would do...

[MigL]

I just love posting videos...

The video quality is bad, but the tune 'rocks'.
( could have posted J Black's version from School of Rock )

Edited November 16, 2020 by MigL