How to improve your data quality from online research

Jim Lumsden

|

24

October 2023

09

January 2022

How to improve your data quality from online research

Jim Lumsden

|

24

October 2023

09

January 2022

Online data collection has revolutionized research. But it does present some new risks when it comes to data quality.

In a lab environment, we can meet our participants face-to-face. We can observe them while they complete the study. Online, things get a bit trickier. How do we know if our participants are paying attention? Are they doing the study properly? Are they who they say they are?

In this article, we’ll explore how to improve data quality in research online by busting bots, cracking down on cheaters, and stopping slackers from spoiling your data. And we’ll explain how Prolific prevent these bad actors from getting into our pool in the first place.

Bots, liars, cheats and slackers…

Malicious participants fall roughly into four groups: bots, liars, cheats, and slackers.

Bots

Bots are software designed to complete online surveys with very little human intervention. You can often spot them by their random or very low-effort/nonsensical free-text responses. Thankfully, because their answers clearly aren’t human, there are several methods for detecting bots in your data (see Dupuis et al., 2018).

Unfortunately, bad-acting humans can prove trickier to detect…

Liars

Liars (or to use the technical term: malingerers) submit false pre-screening information. They do this to access as many studies as possible and maximize how much they can earn.

How badly can this mess up your data? That depends on two things:

  1. How important pre-screeners are to your study design: Let’s say you’re comparing male and female respondents. If some people in each group aren’t who they claim to be, this could invalidate the comparison and you can’t use the data. If sex isn’t a factor in your study design, then it doesn’t matter if a respondent lies about it or not.
  2. How niche your sample is: Impostors are more common when recruiting rare populations (Chandler & Paolacci, 2017). This is because the competition for study places on Prolific is quite fierce. Studies with broad eligibility fill up fast. So, if participants want to maximize earnings, they’ll need to access studies that fill more slowly by claiming to be a member of one (or many) niche demographics.

Cheats

These are participants that deliberately submit false information to your study.

Cheaters don’t always mean to be dishonest. Some may be confused about the data you’re trying to collect, or whether they’ll get paid even if they don’t do “well”. This can happen because they fear your study’s rewards are tied to their performance. They think they’ll only be paid if they get 100% on a test, so they google the correct answers.

Or they might think you only want a certain kind of response (i.e., always giving very positive/enthusiastic responses). Or they might use aides (pen and paper) to artificially perform much better than reality.

The final kind of cheats are participants who don’t take your survey seriously. They might complete it with their friends, or while drunk.

To clarify: Liars provide false demographic information to gain access to your study. Cheats provide false information within the study itself. A participant can be both a liar and a cheat, but their effects on data quality are different.

Slackers

The fourth group are slackers. These people aren’t paying attention and are generally not interested in maximizing their earnings. They don’t feel motivated to give you any genuine data for the price you’re paying.

Slackers encompass a broad group. They could be anyone from participants that don’t read instructions properly, to participants that do your study while watching TV. They may input random answers, gibberish, or low-effort free text.

Slackers aren’t always dishonest. Some may just consider the survey reward too low to be worth their full attention.

These groups do overlap. A liar can use bots, slackers can cheat, etc. Most bad actors don’t care how they earn rewards, so long as they’re maximizing their income!

So, what can you do about it?

Boosting your data quality

We’ve banned our fair share of malicious accounts and learned a thing or two along the way. The tips below aren’t exhaustive but will give you some practical advice for designing your study and screening your data that will boost your confidence in the responses you collect.

Busting the bots

  1. Include a CAPTCHA at the start of your survey to stop bots from submitting answers. Equally, if your study involves an unusual interactive task (such as a cognitive task or a reaction time task), then bots should be unable to complete it convincingly.
  2. Include open-ended questions in your study (e.g., “What did you think of this study?”). Check your data for low-effort and nonsensical answers to these questions. Typical bot answers are incoherent and you may see the same words being used in several submissions (see this blog post for more information and examples).
  3. Check your data for random answering patterns. There are several techniques for this, such as response coherence indices or long-string analysis (see Dupuis et al., 2018).
  4. Looking for a simpler solution? Include a few duplicate questions at different points in the study. A human will provide coherent answers. But a bot answering randomly probably won’t give the same answer twice.

Lassoing the liars

  1. Don’t reveal in your study title or description what participant demographics you are looking for. It may give malingerers the information they need to lie their way into your study.
  2. Re-ask your pre-screening questions at the beginning of your study (and at the end too if it’s not burdensome). This allows you to confirm that your participants' prescreening answers are still current and valid. And it may reveal liars who have forgotten their original answers to pre-screening questions.
  3. Ask questions relating to your prescreeners that are difficult to answer unless the participant is being truthful. That is, if you need participants taking anti-depressants, ask them the name of their drug and the dosage. In most cases, it’d be possible for a liar to look up an answer to these questions… but in our experience, most liars can’t be bothered to invest the extra effort!

Again, we constantly analyze the answer sets of our participants to spot unusual combinations, impossible answers, and other tell-tale signs of malingering.

Cracking down on cheats

  1. Use speeded tests or questionnaires to prevent participants from having time to google for answers.
  2. Ask participants a few questions clarifying the instructions of the task at the end of the experiment (to check they understood the task properly and didn’t cheat inadvertently).
  3. Develop precise data-screening criteria to classify unusual behavior: these will be specific to your experiment but may include:
  • Variable cut-offs based on inter-quartile-range.
  • Fixed cut-offs based on ‘reasonable responses’ (consistent reaction times faster than 150ms, or test scores of 100%).
  • Non-convergence of an underlying response model.

Simple as it seems, it’s been suggested you have a free-text question at the end of your study: “Did you cheat?”

No matter how you clean your data, we strongly recommend that you preregister your data-screening criteria. This will increase reviewer confidence that you haven’t p-hacked.

Sniffing out slackers

  1. Use speeded tasks and questionnaires to prevent participants from having time to be distracted by the TV or the rest of the internet.
  2. Ask participants a few questions clarifying the instructions of the task at the end of the experiment (to check they read them properly).
  3. Collect timing and page-view data by recording the time of page load and timestamp for every question answered.
  4. Record the number of times the page is hidden or minimized.
  5. Monitor the time spent reading instructions. Look for unusual patterns of timing behavior: Who took 3 seconds to read your instructions? Who took 35 minutes to answer your questionnaire, with a 3-minute gap between each question?
  6. Implement attention checks (aka Instructional Manipulation Checks or IMCs). These are best kept super simple and fair. “Memory tests” are not a good attention check, nor is hiding one errant attention check among a list of otherwise identical instructions!
  7. Include open-ended questions that require more than a single-word answer. Check these for low-effort responses.
  8. Check your data using careless responding measures such as consistency indices or response pattern analysis. See Meade and Craig (2012) and Dupuis et al. (2018).

How Prolific picks the best participants

If this seems overwhelming, don’t worry! We’re doing a lot of work on our side to improve the quality of the pool and ensure researchers connect with honest, attentive, and reliable participants on Prolific.

The initial checks

Firstly, we confirm that a participant is who – and where – they say they are. For every person who enters our pool, we verify their:

A participant doesn’t start getting studies until we’ve verified the first three things on this list.  

The ISP and VPN checks

We also check for some more technical things, like:

  • IP addresses
  • The Internet Service Provider (ISP)
  • Device and browser information
  • People using a VPN or proxy

We have a strict list of trusted ISPs. Some ISPs have a high risk of VPN and/or proxy usage. People can use VPNs and proxies to browse the internet anonymously. So, we don’t allow it on Prolific.

It’s critical that our participants are who they say they are.  

The first test

Once we verify a participant, we invite them to complete a test study. Here, we check them for attention and comprehension. The study requires them to write a short story about superheroes.

If the story is meaningful and makes sense, they can do more studies.

The ongoing checks  

When you approve a study, then the participant gets paid and you get the quality data you need. Happy days!

If you reject a study, however, then that gets recorded on the participant’s account. Too many rejections, and they can’t do any more studies.

Feedback

If you’re finding the data from some participants to be unusable, or suspect something fishy going on with duplicate accounts, you can report them to us. You can do this with our in-app reporting function, or via our support request form.

We then review the account and decide if a ban is suitable. Of course, feedback is a two-way street. Participants can also report researchers if they feel they’re not being treated fairly.

These measures ensure you get high-quality data from a panel of the most honest, attentive, and reliable participants.

Getting the best out of the good guys

A critical factor in determining data quality is the study’s reward. A recent study of Mechanical Turk participants concluded that fair pay and realistic completion times had a large impact on the quality of data they provided.

On Prolific, that trust goes both ways. Properly rewarding participants for their time is a large part of that. We enforce a minimum hourly reward of £6 /$8. But depending on the effort required by your study, this might not be enough to foster high levels of engagement and provide good data quality. Consider:

  1. The guidelines of your institution. Some universities have set a minimum and maximum hourly rate (to avoid undue coercion). You might also consider the national minimum wage as a guideline.
  2. How much effort your study requires. Is it a simple online study, or do participants need to make a video recording or complete a particularly arduous task? If your study is effortful, consider paying more.
  3. How niche your population is. If you’re searching for particularly unusual participants (or participants in well-paid jobs), then you will find it easier to recruit these participants if you are paying well for their time.

Error-Free, Confusion-Clear and Engaged

Ultimately, participants are responsible for the quality of the data they provide. But you as the researcher need to set them up to do their best.

  1. Pilot, pilot, pilot your study's technology. Run test studies, double and triple-check your study URL, and ensure your study isn’t password-protected or inaccessible. If participants find errors, they’ll plow on and try to do their best regardless. This may result in unusable or missing data. Don’t expect your participants to debug your study for you!
  2. Keep your instructions as clear and simple as possible. If you have a lot to say, split it across multiple pages. Use bullet points and diagrams to aid understanding. Explicitly state what a participant is required to do to be paid. This will increase the number of participants that do what you want them to!
  3. If participants message you with questions, aim to respond quickly and concisely. Be polite and professional (it’s easy to forget when 500 participants are messaging you at once that each one is an individual!). Ultimately participants will respond much better when treated as valuable scientific co-creators. 🙂
  4. If you can, make your study interesting and approachable. Keep it easy on the eye and break long questionnaires down into smaller chunks.
  5. If you can, explain the rationale of your study to your participants. There is evidence that participants will put more effort into a task when its purpose is clear, and that participants with higher intrinsic motivation towards a task provide higher-quality data.

What to do if you think your data quality is compromised?

Firstly, talk to us. We’ll always ban participants using bots or lying in their pre-screeners. You should reject submissions where you believe this to have occurred and send us any evidence you’ve gathered. Data quality is our top priority, so please reach out to us if you have any concerns, queries, or suggestions.

In cases of cheating or slacking, we ask that you give participants some initial leeway. If they’ve clearly made some effort or attempted to engage with the task for a significant period but their data isn’t good enough, then consider approving them, but excluding them from your analysis. If the participant has clearly made little effort, failed multiple attention checks, or has lied their way into your study, then rejection is appropriate. Please read our article on valid and invalid rejection reasons for more guidance.

You can also learn more about how to improve data quality in research in The Complete Guide to Improving Data Quality in Online Surveys. You’ll discover:

  • Common data quality issues
  • Tried and tested ways to improve the data quality from your online survey responses
  • How Prolific can help you achieve high data quality

Download your copy now.

Leave your details below and someone will get back to you