08/28 - Tranquility Tuesday!

Did anyone ever read O.J. Simpson's "If I did it"?

Not asking for any reason in particular.

If I were to write a bot, though...

Seems like there's a whole bunch of FUD. I'm trying to engage some requesters in dialog about it but you get you're screwed, no matter what, right?

Let's talk batches. Let's say some naughty person were to completely script them. Totally against TOS. How would you, as Amazon or the requester go about detecting it? If the data submitted for that batch is completely accurate from a bot... maybe the submission time? What if that were cooked? Guess who'll look like a bot? The normal user who set up keybinds or even a rudimentary script to pull in data! [Think of the old A9s or similar where you needed the product page.] The botter? They've cooked their times. #WalterWhite

But, you say, you can detect humanity with well-done mouse-y look-y-loo. Sure. In fact, I think this is part of what Google's whole reCAPTCHA does -- look for patterns of mouse movement that's distinctly human.

Would this make sense for, say, the A9s that are currently up? If someone could do a bot for them, first, holy crap, welcome to the singularity, second, Amazon's going to stop putting 'em up.

The fact is, you can't "bot" most batch hits. Some, sure. The requester doesn't know rudimentary anything, and it's a great boon for workers who can whip together something *potentially* *cough cough* TOS violate-y.

But guess what? That requester's gonna get the same data if it wasn't botted. The data will be correct. The worker didn't manually put a link into their browser, select a portion of text with their mouse, copy it, go back to the previous tab, paste it, and click on the submit button, but they've sent the same data as if they had. I'm fairly certain this isn't the 'problem' that's scaring requesters.

It's all 'bout the surveys. [Weird Al or Meja here as preference dictates.]

So think about it. How do you bot a survey? Actually, how do surveys work in real life? You go into a room (one of you). If the researcher really cares, they'll check your ID to make sure you, are you. You start filling out those bubbles. You turn it in. You get paid (hopefully!) and leave.

How do they stop people from 'botting' those?

Ya'll ready for this?

Internal consistency!

"I hate everybody and never leave my house" [1] [2] [3] [4] [5]
"I love partying and go out every chance I get, because I love people!" [1] [2] [3] [4] [5]
If someone checks 5 for agree on both of those, they just very well might be randomly answering stuff.
Like a bot.

Don't we see those on surveys a lot? Is that less effective digitally? Nope.

It's frustrating me just how speculatory things are getting, like, someone being banned for using that random extension. Sure, it's a non-nil probability, but you've got to expect lots of people installed it while publicly decrying it. The more likely pattern is that some requesters have bought into this 'bot' hype, and without thinking "Hey, we can check their IP and look at their responses for internal consistency", they're instead looking and saying "Completed this page in x.y seconds? Bot!" and noting it to Amazon and Amazon's all like, "Ban."

The guy that started this was a moron. He's new to the platform and he's lazy. Instead of learning how to use it and conduct research responsibly and productively, he complained the sky is falling. Because there isn't any other side out there (note there isn't a follow up Wired article to the ending of "the jury still being out") more requesters are buying into this without stopping to think, "Wait, we already have means of preventing this."

Want to try to help get things back on track? Send messages to your favorite requesters, especially if you're big into surveys. Let them know you appreciate their research and that you're there to address concerns they may have about this "bot scare" they may have heard about.

You wouldn't have to point out to someone like rowan that if it were possible for a bot to do his hits, why, that would open up a whole new and fascinating aspect of research. It's almost certain that anyone working on anything AI related would love to hear about any automated thing that can outperform a human. (Spoiler: Nothing can yet.)

Let's try to educate our requesters so that the fake news (teehee) out there isn't the only thing they've got.

But also, still bombard the heck out of Joe Lo because of his [object object] broken piece of [object object] server that [object object] on the [object object].

 

Title: Session 1/4 cognitive task and survey experiment. MICROPHONE REQUIRED(~ 60 minutes) [ $ | $ ] PANDA
Requester: CCP Lab [A1WSKAJH3W4OEL]
TurkerView: [ $10.01 / hour ]
Description: Vocal color naming task with surveys. Compensation $5.00. Approximately 30min. Requires microphone, Google Chrome, and Windows PC.
Duration: 04:00:00
Available: 0
Reward: $5.00
Qualifications:

• Total approved HITs GreaterThanOrEqualTo 100,
• TP Panel: -19308359 N GreaterThanOrEqualTo 100, [ Request | How to Use ]
• HIT approval rate (%) GreaterThanOrEqualTo 10,
• Exc: [1919008766-103998] DoesNotExist [ Request | How to Use ]

[3AX4X2ATGIHOC9CKXQT4IS0APNAYDN]​

 

SurveyComet [A3NC2P2SPK9QQ] College Alcohol Study (≈ 60 minutes) - $10.00 | PANDA

[3KNJI1FZ8L9Y31XCQC6O1DB77J3QNQ]​

Generous

Unrated

Unrated

$26.30 / hour
00:22:49 / completion time

​

Pros:

Cons:

To read SurveyComet's full profile check out TurkerView!​

 

Title: Webcam survey - New HIT posted on August 20, 2018 [ $4.89 | $4.89 ] PANDA
Requester: Michael Seiler [A38TNDDF5JITK8]
TurkerView: [ $6.97 / hour ]
Description: You MUST have a WEBCAM and have owned a home in the US
Duration: 00:00:00
Available: 1
Reward: $1.06
Qualifications:

• HIT approval rate (%) GreaterThan 95,

[3CL9ACF20KN7P2D1KC74ST7SDRO6W4]​

 

Title: Webcam survey - New HIT posted on August 20, 2018 [ $4.89 | $4.89 ] PANDA
Requester: Michael Seiler [A38TNDDF5JITK8]
TurkerView: [ $6.97 / hour ]
Description: You MUST have a WEBCAM and have owned a home in the US
Duration: 00:00:00
Available: 1
Reward: $1.06
Qualifications:

• HIT approval rate (%) GreaterThan 95,

[3CL9ACF20KN7P2D1KC74ST7SDRO6W4]​

 

Amazon Requester Inc. - A9 Data ValidationSleeve Length (WARNING: This HIT may contain adult content. Worker discretion is advised.)$0.01Approved From yesterday ( I wish I would have did more of these!)

Yet:
Amazon Requester Inc. - Customer InterestsProduct to Interest Audit (single yes/no per HIT)$0.01Pending from Friday. I thought these approved within 48 hours?

 

ugh, I've been using Panda Forker and saying show only those which I am qualified for. I turned it off today. There is so much work I am not qualified for D: I can't even do some A9s right now

 

ParkLab [A19KFNEFOL40ND] Answer a survey about your opinions - $2.00 $2.00 bonus | PANDA

[3ZJNQ0RXZJVK9J64RLXUQ5EBFF3XXC]​

Fair

Unrated

Unrated

$8.83 / hour
00:27:10 / completion time

​

Pros:

Cons:

To read ParkLab's full profile check out TurkerView!​

 

"panda forker" gave me a great mental image of someone poking at @A6_Foul_Out with a fork.

 

Those are not the A9 I want.

 

How do I get the str11223344 qual?

 

Almog Simchon [A263XRXWQXBFVP] ~3 minute scenario judgment(~ 3 minutes) - $0.36 | PANDA

[3SZ4WBITUE9PLRMSQW0QX1XUWDXXKA]​

Unrated

Unrated

Unrated

$10.20 / hour
00:02:07 / completion time

​

Pros:

Cons:

To read Almog Simchon's full profile check out TurkerView!​

 

Linda Hagen [A363HPIHVFOR0M] Rate a food consumption situation (1-2 minutes) - $0.20 | PANDA

[3OSG0B9IUL0UN5D6Y1D3DAQHIJ8IVW]​

Unrated

Unrated

Unrated

$6.86 / hour
00:01:45 / completion time

​

Pros:

Cons:

To read Linda Hagen's full profile check out TurkerView!​

 

It hasn't been open in years.

 

I think the C-STATS are better than the A9's today. I hate the color ones.

 

Go back in time

 

fail, since they didn't include:

"A horse walks into a bar and orders a beer. The bartender asks, "Why the long face?"