We needed to pay people small sums for crowd sourcing tasks, mat purchase and other tasks but we also wanted to explore putting distance between our accountant and the payee, by transferring money through an unknown (and potentially untrusted) intermediaries. The question, in a nutshell — if Ai pays Baozhai and asks him to forward the money to Chunhua, what is the likelihood of the money arriving in Chunhua’s account?
Again we posted a question to the Sandaha crowdsourcing site asking for volunteers to receive and pass on 50 RMB, for the opportunity to receive a 2.4 RMB reward once receipt has been confirmed. The task was initially rejected by Zhu Ba Jie for being too sensitive, we assume because it deals with anonymous money transfers, something that is an issue the world over (for those interested in the topic, read the Know Your Customer, KYC requirements that financial institutions need to adhere through). With adapted text, the task was accepted on Sandaha where 161 persons paid attention to the task, 13 signed up to be part of the experiment and 10 people were selected to be intermediaries for the money. In the public forum the members that responded to this task questioned its purpose assuming that it was one way of finding team members that can be trusted for an apparent follow on task; and it’s a test of people’s moral quality. In a follow up question we asked whether they would return a wallet if found on the street, most said no unless they thought it was part of a TV programme and they were secretly being recorded.
We received confirmation that 6 final recipients (60%) received the payment, implying that 4 intermediaries kept the cash (there are other variations, the final recipients could have lied about receiving the final payment in order to entice a second payment, the final recipients who ‘ received’ money could have been lying for some unknown reason; money could have been sent by the intermediary but was not received by the recipient; or the intermediary and recipient could collude or even be the same person — fairly standard cryptography challenges to solve).
It might be argued that the 40% “failure rate” is a successful filtering exercise where (in the context of this single task) 4 people removed themselves from pool of people for whom a follow task might be run. Interesting follow up tasks include repeat payments through one individual, and reactions to different cues as to the purpose of the payment.
While this is an interesting research topic in its own right, we decided to proceed using direct payments between the accountant and the recipients.