An upcoming alternative review platform for mTurk, which fundamentally rethinks how reviews on the platform should work.
Get quick, skimmable updates from people whose judgment you trust, rather than meaningless aggregate scores from people who you have very little in common with.
Review platforms work best when they are:
- Simple (rather than complicated)
- Exclusive (rather than unchecked)
- Reciprocitive (rather than parasitic)
- Current (rather than outdated)
- Responsive (rather than defunct)
- Free (rather than expensive)
Most other mTurk review platforms operate on a 5-star review platform with 4 or more separate criteria.
To reduce the amount of time it takes to complete or read a review, ThumbWork will reduce this to 1 simple criteria, with two possible options:
- Good 👍
- Not Good 👎
Other platforms also encourage some users to wax poetic on their soapbox, writing an entire essay about how they were wronged. Or perhaps they're writing a long review on the nuance of how this HIT is different from others. It's almost as though they're sipping fine wine and asking you to smell the bouquet, and it probably belongs on Tumblr rather than on a review platform.
We're here to work. Tell me quickly whether I should do a HIT, or not. I have things to do.
ThumbWork will allow only 80 characters per review, encouraging reviews that are quick to write and easy to skim.
(A positive side effect of their short length is that they will be very easy to embed in various locations.)
People have wildly different expectations, and the differences between people make most feedback useless to most other people.
The current popular platforms operate on a (clearly-flawed) assumption that everyone's opinion is equally worthwhile. From racist trolls to clueless noobs, there are plenty of counterexamples available on an hour-to-hour basis showing that this isn't the case.
In the real world, you can't trust anoymous aggregate information. The source of your information is usually more important than its contents.
ThumbWork will operate on a "friends" paradigm: People will see only the reviews of people they are "friends" with.
It's easier for small groups to enforce reciprocity, keeping the entire group more efficient as a whole by ensuring that nobody leeches off the hard work of only a few.
ThumbWork's "small group" approach should help ensure that no member is dead weight.
Information loses value as it ages.
Often, information that has gone sour with age can be harmful, just as eating rotten food when you're hungry can make you sick.
Reviews on ThumbWork will last a maximum of 24 hours, in order to better highlight recent information that's still within its shelf life.
When an underfunded service needs to store and retrieve excessive amounts of data regularly, it starts to suffer performance and availability issues:
Example of an unresponsive, indrectly government-funded service:
A service that becomes unavilable as a consequence of trying to do to much is entirely worthless.
ThumbWork will stay responsive by restricting its scope and making it very easy for individuals to self-host review groups with a lightweight, easily-configurable server.
Without re-thinking the fundamental paradigm, the only way to provide a service that sends too much worthless information to too many people is to start charging for it:
Example of an expensive, full-featured review service:
While you can't fault the platform for charging to cover the costs of the service that it offers, it really begs the question:
"Is all of this really necessary?"
Because ThumbWork will be open-source and will put hosting in the hands of users, there will be no inherent need for a centralized authority to charge users for the opportunity to use it.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent