Beyond the Turk: Alternative platforms for crowdsourcing behavioral research
Abstract
The success of Amazon Mechanical Turk (MTurk) as an online research platform has come at a price: MTurk has suffered from slowing rates of population replenishment, and growing participant non-naivety. Recently, a number of alternative platforms have emerged, offering capabilities similar to MTurk but providing access to new and more naïve populations. After surveying several options, we empirically examined two such platforms, CrowdFlower (CF) and Prolific Academic (ProA). In two studies, we found that participants on both platforms were more naïve and less dishonest compared to MTurk participants. Across the three platforms, CF provided the best response rate, but CF participants failed more attention-check questions and did not reproduce known effects replicated on ProA and MTurk. Moreover, ProA participants produced data quality that was higher than CF's and comparable to MTurk's. ProA and CF participants were also much more diverse than participants from MTurk
Study specs
- Authors
- E Peer,L Brandimarte,S Samat,A Acquisti
- Institution
- Eller College of Management,University of Arizona,Heinz College,Carnegie Mellon University
- Discipline
- Behavioral Science
- Year
- 2017
- Human Data Platform
- Prolific
- Source
- View Source Google Scholar
Peer Review & Critical Discussion
Potential Selection Bias in 2023 Cohort
The participant pool shows a concerning overrepresentation of users from high-income demographics. Looking at Table 3, we can see that 78% of respondents had annual incomes above $75k, which significantly limits the generalizability of these findings to broader populations.
Non-naive Participants Issue
I've noticed a methodological concern regarding participant naivety. Given that Prolific users often complete multiple studies, there's a real risk that participants had prior exposure to similar experimental paradigms, which could confound the results.
RLHF Applicability to This Study Design
The implications for RLHF training pipelines are understated. If we accept the authors' conclusions about preference stability, this has direct consequences for how we should structure reward model training. The temporal decay effect described in Section 4.2 is particularly relevant.
Verify your expertise to join discussion
Create an account and verify your credentials to participate in peer discussions.