New Gun Myths


#1

Continuing the discussion from HYST Generic Discussion Thread:

I’ve created a twitter bot inspired (in response to?) the NRA’s gun-related Fairy Tales:

http://www.npr.org/sections/thetwo-way/2016/03/25/471726704/nra-rewrites-fairy-tales-with-more-firearms-less-bloodshed

 

aaaand, first iteration is live:

#https://twitter.com/NewGunMyths


Another iteration went up last night, killed the app overnight, fixed the app (needed to update a config setting that worked locally, but not in the cloud - DOCUMENTED FOR THE FUTURE), aaaand… it’s not really all that visible a change.

I’ll have to think about this some more.

When I dump out 100 phrases locally I immediately spot lots of teh funny. Not so when it’s only once per hour.

However, a lot of the work I’ve been doing has been to make things more modular and add tests – so I can now iterate a bit faster and securely. HOPEFULLY.

As usually, all of this is boring stuff nobody cares about.


#2

Does it learn from things we tweet at it?


#3

Ooooh, wouldn’t that be a useful addition?


NO IT DOES AND WILL NOT


Coincidentally, @LDoBe and I were talking about that issue in the Huffing thread earlier this year:


#4

Please let it learn from my tweets. It’s the only thing worth tweeting for.

And can it support learned regexes?


#5

Hrm. Maybe it’s getting better:

Must be learning from all of those pro-gun tweets it’s been getting from @daneel.


#6

https://twitter.com/daneel?

I can’t understand that guy at all.


#7

Not going to incorporate any of this?


#8

Thanks! I went to a botmakers slack (that I’ve been mostly low-level lurking on) to see their reactions, and discovered the author of the NodeJS portion of the BotFramework had just joined up!

I’ll look at it later. I think the thing is mostly for interactive bots - which may be “the future” but it’s not my bot area at the moment.


Today’s output looks a bit better, but there’s still too much repetition. The app is running on Heroku’s free plan which requires 6 hours of downtime a day, so is triggered once an hour by their scheduler. Which mean no saved state - so instead of randomly cycling through a diminishing array (pick and remove) it randomly picks from the full array each time. It’s a small array, and the random generator seems to specifically want to annoy me.

&nbsp

 


#9

Wouldn’t it just be easier to train it to retweet hollywood studios?

How does it make the NRA feel that the hollyweird left coast had them beat by years in this little game?


#10

That would be @RealHumanPraise:


#11

Here’s why bot-ethics is a thing. This tweet could be very disturbing for many people:

It was merged from the two following sentences (the first is an Arne-Thompson myth class, the second is from my list of gun-related phrases):

Jealous co-wife kills woman's children.
Guns kill.
=> Jealous guns kills woman's children.

A lot of bot-ethicists would say to avoid such things entirely.

But since I specifically dealing with guns, such troublesome combinations are an expected result.


#12

I thought there was one rule three laws when it came to bot ethics.

Please tell me your bots obey the three laws. Or at least they don’t talk about it.


#13

Taboo, not Tabu…unless you’re talking about the Indian actress, or the movie.


#14

The “myth” components, including the spelling of “tabu”, come from the Arne-Thompson classification

S. Thompson. Motif-index of folk-literature : a classification of narrative elements in folktales, ballads, myths, fables, mediaeval romances, exempla, fabliaux, jest-books, and local legends.

Revised and enlarged. edition. Bloomington : Indiana University Press, 1955-1958.

http://www.ualberta.ca/~urban/Projects/English/Motif_Index.htm
http://www.ualberta.ca/~urban/Projects/English/Content/c.htm


#15

How long did it take you to find that reference? :wink:


#16

Not long - it’s part of my notes, since that’s where the bulk of the text comes from.


#17

However, it can be replaced. Something I’ll think about. Thanks for the feedback.


#18

#19

Well, I’d suggest that an obscure 1950’s reference is not the one to use if on top of being wildly out of date it’s also causing you to spell a word incorrectly. It makes you look like you don’t have good information. That’s the opposite effect of what you’re trying for!


#20

In the field of narratology it’s hardly obscure.

Although one could argue it’s an obscure field. Then again, I’m making twitter bots. Obscure piled on Obscure piled on Obscure.