08/08 - Thickened Thursday!

NO don't do this

wearing gloves shows intent....or so I've heard

don't ask

 

ok, wear your neighbor's gloves too

 

The sad thing is, I'm not pissed he was getting high, college kids do that. I'm pissed he was using it and storing it in my fucking house after he was told specifically not do that.

 

Title: Evaluate the quality of two short speech recognition transcripts | PANDA
Requester: Anya Korneyeva [A14IX62ZL35E49]
TurkerView: [ $ / hour ]
Description: Listen to a short audio recording, read two speech recognition transcripts, tell us which one is better
Duration: 1,440 Min
Available: 1
Reward: $2.50
Qualifications:

• Location In GB, IN, US,
• HIT approval rate (%) GreaterThan 95,

[3EWXWOT88PDU3563ZW5XE85G7VIP5U]​

 

Title: Which Conversational Partner is Better? | PANDA
Requester: Noah Turk [AHJQF4WEPBEHJ]
TurkerView: [ $14.37 / hour ]
Description: Evaluate quality of conversations through comparison. (ID: pairwise_dialogue_eval_1565120286)
Duration: 10 Min
Available: 6
Reward: $0.50
Qualifications:

• Location In AU, CA, GB, NZ, US-AL, US-AR, US-DE, US-FL, US-GA, US-IA, US-KS, US-KY, US-LA, US-MD, US-MN, US-MO, US-MS, US-ND, US-NE, US-OK, US-SC, US-SD, US-TN, US-TX, US-VA, US-WV,
• pairwise_dialogue_eval_1565120286_max_submissions DoesNotExist
• pairwise_dialogue_block_onboarding_fails DoesNotExist

[3OEIII5REGSD745ZTTMYMGWT5KEMSP]​

 

Johnson School BSL [A708YWZMMMGCY] Short survey about grocery shopping(~ 9 minutes) - $0.90 | PANDA

Generous

Unrated

Unrated

$19.76 / hour
00:02:44 / completion time

​

Pros:

Cons:
• Unpaid screener

Qualifications: SurveyGroup [2706] DoesNotExist ; Total approved HITs >= 100; HIT approval rate (%) >= 95; Exc: [2111164845-163537] DoesNotExist ; Location In US;

To read Johnson School BSL's full profile check out TurkerView!​

 

FWIW TV had a 70/30 split last month of "professional" Turkers vs what would probably be deemed as more casual/normal folks. It's probably even closer to parity if you consider full-time workers who don't blaze through things.

One of the more frustrating experiences for the faster workers is this watering down of the userbase (from their POV) and there has been infinite levels of conflict&complaints as those users move into the minority same as excessively slow users feel the pain.

Which is why as TV has matured it's made more of an effort to display compartmentalized groupings of data, and I generally will throw specific concessions to that group of workers because they're representative of what I'd like MTurk to look like for everyone, and I appreciate their cute faces.

& truthfully, I need to find ways to do a lot more for those folks because they're the ones most hurt by lapses in TV's caring for them.. but since they're also the group who tends to be experts on MTurk, it's relatively difficult to actually do that lol. Hard to help someone who can help themselves (& is the cause of this phenomenon).

Thank you. This is basically the best reply to a question framed only considering one side, with the addendum they are filtered out of aggregate datasets. They're not "deprecated" and will be displayed where appropriate (mostly the website, though given the complex nature of the # of filters on there I KNOW there are bugs w/ that)

It also just tends to happen less for overly fast workers because people going at that pace probably aren't completing the work in good faith to begin with, and have absolutely no reason to contribute to nothing more than a morality based help guide. It's a self solving problem, whereas the inverse group has (as a generalization) different incentives/reasons for 1) completing work at the pace they do and 2) broadcasting that out to a platform that, at its core, "sides" with the idea that low wages should be increased. For them, leaving a review is self-serving (not in a malicious way, I don't believe most intend to do harm to other participants on the platform). They want to be represented, which is why TV only excludes them from worker facing datasets but will still display them on the website for visitors to consume at will & make their own judgement on, because they can do that through that medium.

Also, from a technical perspective, the act of Turking lends itself to more lower-bounds skewed data than upper-bounds. You have to put forth excessive amounts of effort to reach the upper bounds, whereas you just have to pee really bad in the middle of a survey to throw your time into the toilet (ayooo)

RA is basically the upper threshold benchmark of what I'd let continuously be submitted. There are an equal # of people who take exception to that threshold as those who take exception with the lower bounds threshold. <spider-man-pointing.pixelatedJPEG> & I understand and regularly field arguments from both sides, not a problem since it helps gauge where humans are at with reacting to TV.

But there is no "ground truth" to all this. If someone spends 10 minutes reflecting about their 50 years on the planet during a "do you feel alert/aggressive/depressed/hungry" battery their time is not incorrect. If someone, intending to display that the survey manipulations were utterly ineffective, straightlines that same questionairre in half a second their time is not incorrect. Something both extremes just regularly refuse to accept is that the requesters actually appreciate both of those responses (to a point). Brace yourselves, and take a seat if you use a standing desk, but 99% of us are terrible social scientists!

 

hey this belongs to @THFYM not u

 

@Ten your dad is talking about kicking you out again.

 

reading hard

 

tl;dr: slow the fuck down, cyborg.

 

New Requester Profile!

Anya Korneyeva [A14IX62ZL35E49] Evaluate the quality of two short speech recognition transcripts - $2.50 | PANDA

Generous

Unrated

Unrated

$37.19 / hour
00:04:02 / completion time

​

Pros:
Easy enough, good pay

Cons:

Qualifications: Location In GB, IN, US; HIT approval rate (%) > 95;

To read Anya Korneyeva's full profile check out TurkerView!​

 

I wouldn't uninstall MTS. The Hit Tracker is very useful. Plus the other scripts it includes like workspace expander, etc. Just ignore the finder/catcher and use forker/panda crazy instead for that part. MTS is still useful even if you don't use the finder/catcher built into it.

 

CoCoSci Lab at Princeton [A1S8PU3QK4OKG3] Solve anagrams (takes ~5 minutes) - $1.25 | PANDA

Generous

Unrated

Unrated

$21.13 / hour
00:03:33 / completion time

​

Pros:

Cons:

Qualifications: Total approved HITs > 100; Masters Exists ; HIT approval rate (%) > 95; Location EqualTo US;

To read CoCoSci Lab at Princeton's full profile check out TurkerView!​