There is no excuse for treating domain experts like bureaucratic liabilities, and SnorkelAI's handling of this paid survey is a textbook case of how not to engage the research community. A machine learning researcher with a PhD and A* publications was contacted, verified, and then subjected to an experience that was, by their own account, pushy, disorganized, and ultimately insulting. The takeaway for our readers is simple: if your data collection process demands more paperwork than a public-sector contract and requires a personal visit to a tax office, you are not running a survey, you are running a gauntlet.
This story matters because it sits in direct tension with the kind of thoughtful, human-centered data management we advocate for. Compare this experience to the measured, citation-driven influence we saw in Anthropic's research influence outpaces OpenAI's in citation rates. Anthropic earns credibility through substance and community respect. SnorkelAI, by contrast, appears to have treated a qualified researcher as a checkbox in a broken workflow. The researcher reported receiving follow-up emails every two to twelve hours while their own replies went unanswered for days. That asymmetry, aggressive automation paired with slow human response, signals a system designed for the company's convenience, not the participant's dignity. And when the payment process required documents that take up to four weeks to obtain in person, the company effectively turned a paid engagement into an unpaid errand.
For researchers who have navigated the grueling peer-review process described in Navigating Peer Review for Original Classical ML Research at Top Conferences, the contrast is instructive. You will fight for months to get a paper accepted, only to be treated as an afterthought when a company wants your expertise for market research. That is not a partnership. It is a transaction where the terms are hidden until after you have already invested your time. The researcher's recommendation, "run away", is blunt, but it carries the weight of someone who knows what professional respect looks like and knows this was not it.
The concrete point to watch is whether SnorkelAI learns from this feedback or doubles down. The researcher's account is not an outlier complaint; it is a documented failure of process design that directly undermines the company's ability to attract the very experts it needs. Any organization building AI tools must recognize that the people who can best evaluate those tools are also the people who are least likely to tolerate poor UX, aggressive follow-ups, and punitive administrative hurdles. If SnorkelAI wants to be taken seriously by the research community, it needs to start treating participation as a collaborative exchange, not a bureaucratic burden. Until then, the smartest move for any researcher contacted by them is to read this account carefully and decide whether their time is worth more than the promised payout.