Data quality in online human-subjects research: Comparisons between MTurk, Prolific, CloudResearch, Qualtrics, and SONA
Abstract
With the proliferation of online data collection in human-subjects research, concerns have been raised over the presence of inattentive survey participants and non-human respondents (bots). We compared the quality of the data collected through five commonly used platforms. Data quality was indicated by the percentage of participants who meaningfully respond to the researcher’s question (high quality) versus those who only contribute noise (low quality). We found that compared to MTurk, Qualtrics, or an undergraduate student sample (i.e., SONA), participants on Prolific and CloudResearch were more likely to pass various attention checks, provide meaningful answers, follow instructions, remember previously presented information, have a unique IP address and geolocation, and work slowly enough to be able to read all the items. We divided the samples into high- and low-quality respondents and computed the cost we paid per high-quality respondent. Prolific ($1.90) and CloudResearch ($2.00) were cheaper than MTurk ($4.36) and Qualtrics ($8.17). SONA cost $0.00, yet took the longest to collect the data.
Study specs
- Authors
- BD Douglas,PJ Ewell,M Brauer
- Discipline
- Social Science Research Methods
- Year
- 2023
- 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.