08/31 - Striking Saturday!

Someone called?

 

Was out looking down to see the 3 stray/ferals I feed and saw a teeny weenie black new kitty scoot off from the food dishes Need to catch the lil bugger tomorrow.

 

I think anyone's answer to this will mostly boil down to opinion, some think the current system is fair (the ones who put in effort to maximize its returns) & others don't. Kind of the bigger "is this possible" problem I'm not qualified for

100% agree with this. It's why I thought the idea in a design sense was interesting. MTurk is (probably) a little too wild west for the workers at expense of requester's needs (demographic quals are INSANELY overpriced & it hurts everyone, nvm the rest of the design shortcomings) whereas Academic designed systems usually swing too far in the other direction (my opinion). Ditto for TurkPrime's proposed "ideal" qual system if you read their paper. Their functional platform strikes a solid balance though, tending to safeguard against people who are wildly ineligible to complete work but still allowing enough freedom for everyone else to figure out their own thing. They could probably improve on some points, but since they're stuck w/ the framework they operate in, I think they do really well.

Although for MTurk specifically: I think technically polling isn't bad because their business is web services, and polling is cheap (at least relative to what they charge since all the typical business costs are in-house'd). The time/cost spent redoing the system to support this might not actually pay a worthwhile ROI. Similar to requesters you'd think would leave & go do their own thing but don't because it's just cheaper to not "fix what ain't broke" sort of thing - if that makes sense?

FWIW, I think PC's system is equally bad.

1. It interrupts & distracts the worker (whether the interruption is PC's insane "NEVER TAKE PRESENTS" alarm or a modal requiring user interaction to indicate availability to be assigned the HIT - both are bad) OR
2. It allows assignment of HITs to users unavailable to complete them (leaving pandas on and going AFK, or a system which has no user confirm to assign in a pub/sub). Anytime either system assigns or allows to be assigned a HIT to a worker unavailable/ineligible to complete the work it comes at the expense of another worker. In an active system (PC) the worker is at least always managing a very small pool (2-3) of HITs they can even be eligible for in a trade-off situation. In a passive (pub/sub) I feel like it is more difficult for the larger system to function efficiently in regards to active/inactive||eligible/ineligible clients. If the system steps in, well see the point about ICs/liability

If you fix #1 you tend to have a problem with #2. If you want to fix #2, you're going to have to start ditching user privacy to track their focus/attention/availability which I've seen done but it's kinda sketch (just an opinion here, ofc). There is an article somewhere I'll never find about a guy who has his own like digital workforce and the software completely "owns" the system, users are tracked hardcore and it's mostly exploiting the kind of labor you'd expect to sign on for that (3rd world digital labor) but it's interesting nonetheless from the "ok how do we solve #2" perspective.

FWIW, I find the idea legitimately interesting

It's relevant to my interests and core to something I'm working on now. I'm not arguing because I think you're wrong, I'm just arguing to find finer details in how it'd work/not.

Appreciate the idea. I've seen system floated to be more "fair" but w/ zero regard for the technical portion. Yours was better from a tech perspective b/c I am incapable of solving the fair portion LOL

 

A couple years ago, Laura recognized me at a show in Bloomington because I'd already been at shows in other cities. Oops.

 

It's easy though to change the settings in PC to use different sounds.
(I just use the little "clunk" sound.)

 

The mturk server is hilariously overkill. From what I gather it is actually comparable to amazon proper.

So yeah I don't think that they care.

 

I pray for Chaos, Gus brings Chaos, I pray for Gus.

 

RIP you if you're stalking her

 

Team Chaos!

 

No, it was just an unprovoked wave when she looked over at me lmao.

 

I pray for tacos.

 

I've read academic research papers with less thought put into them.

Maybe that's not saying much... but still, I've read academic research papers with less thought put into them.

 

@morweeg

 

@jessers

 

@taco

 

Whole different problem.

Maybe NSF grants should be paid out according to MTurk guidelines.

Conduct the research first, submit it, and if it's >=99% accuracy you can have the $. Otherwise, FOH.

 

This seems like a problem for smarter minds. I'm too busy trying to edit TVJS code to hide hits from people that don't pay us $49.99/month.