Richard Mills

ManipulationDetect: An AI Auditing Tool for Online Choice Ar Policymakers and regulators are increasingly interested in behavioural auditing tools to counteract manipulative designs in Online Choice Architecture (OCA). To date, auditing tools have been largely

Thrilled to share a new paper with @stuartmills.bsky.social &
@casssunstein.bsky.social. The paper discusses ManipulationDetect (previously NudgeDetect), a tool I launched a few months ago that uses LLMs to highlight 'manipulative' practices online.

ideas.repec.org/p/not/notcdx...

Wow, I didn't know I've authored an autobiographical memoir...

"I am the receiving editor of a {XXX} journal, and after reading your recent publication "Mills, Richard", I was truly impressed by your outstanding academic contribution."

I must say, it is a riveting read,

Thank you to all my wonderful co-authors. You can read the paper here: doi.org/10.1037/hea0...

What drives cooperative health behaviors? Our study of 3,000+ UK individuals during COVID-19 found past prosociality (e.g., volunteering) was a key correlate of adherence, but only when rules were 'ambiguous'. Prosocial people also vaccinated to protect others ("We"), not just for themselves ("Me").

Thanks @koenfucius.bsky.social - appreciate the support! I've been working on a better version (which should be public soon)

P.S. When you install the extension, Chrome will say it “can: Read and change all your data on all websites.” This is a standard warning. In reality, the only thing accessed is the webpage’s visible text, which is sent to the AI for analysis. No personal data is ever stored.

Thank you so much for the support here too @dilipsoman.bsky.social 🙏🏼

Thanks for the support! @ajohnstone.bsky.social

The plan is for it to be used as a research tool (though still fairly early stages in its development - just an initial proof of concept)

It’s free (up to 30 scans p/m) and totally anonymous.

This tool is still very much in early stages, with some bugs – but any and all feedback welcomed.
Let’s build a more honest internet.

Happy NudgeDetecting!

🟢🟠🔴

5/5

Examples of what it catches:

• FOMO (“Only 1 left!”)
• Anchoring (“Original price £1,105. Current price £942”)
• Hidden Costs (Drip pricing) ("Concert tickets from £20, final checkout price £37.50 after booking fees")
• Upselling (“Buy a bundle”)
• ...and many more

4/5

Why I made it:

The internet is full of sneaky design choices that often push us to spend more, share more, or act against our own interests.

NudgeDetect helps you see through these tricks.

3/5

What NudgeDetect does:

1️⃣ Scans any webpage
2️⃣ Detects up to 5 behavioural nudges, sludges or dark patterns
3️⃣ Highlights them directly on the site
4️⃣ Rates their severity (🟢 Low | 🟠 Moderate | 🔴 High)
5️⃣ Explains + suggests how to resist

2/5

🚨 NEW TOOL ALERT: Meet NudgeDetect 🚨

Ever felt nudged into clicking "Buy Now", unknowingly subscribing, or paying hidden fees?

NudgeDetect is a Chrome extension I’ve built to reveal ‘manipulative’ website tricks: and help you fight back.

chromewebstore.google.com/detail/aodlb...

1/5

A pleasure to present “AI in Action: Practical Applications for Researchers” with the brilliant @muskaanp.bsky.social at yesterday’s PGR Conf @notts-psych.bsky.social. We covered LLMs for thematic analysis, web data analysis (NHSBT), and building AI tools. Thanks all who joined and stayed to chat!

It was a real pleasure (and honour) to be a part of this, writing a short chapter on 'Choice Bracketing' with @danielread.bsky.social

Thanks to ChatGPT for being the brains behind this operation. If anyone is interested in how I set this up and would like to implement it in their own work, I’m happy to share more details.

P.s. the Voucher codes aren't real... :D

#Qualtrics #ResearchTools

Once assigned, the script marks the voucher as “Used” (with a timestamp) to prevent two participants receiving the same code. The script is deployed as a Web Service, which Qualtrics calls in real-time (extracting the voucher code and displaying it to the participant at the end of the survey).

After digging through the Qualtrics community pages, the best solution I found was using a Google Apps Script that retrieves a unique, unused voucher from a Google Sheet.

For those using #Qualtrics, this might come in handy one day! 🚀

In one of our projects, we needed to provide participants with a unique and anonymous gift voucher upon survey completion. The challenge? Qualtrics doesn’t have a built-in way to randomise and distribute unique voucher codes.

If it turns out LLMs are only capable of recombinatory innovation (finding novel connections among existing knowledge), that would still be very useful. Most innovation is recombination and one of the big issues in science is that fields are too vast for scientists to bridge them to find connections

Generative AI has flaws and biases, and there is a tendency for academics to fix on that (85% of equity LLM papers focus on harms)…

…yet in many ways LLMs are uniquely powerful among new technologies for helping people equitably in education and healthcare. We need an urgent focus on how to do that

Prophecies of the Flood What to make of the statements of the AI labs?

There has been a definite shift in recent weeks where insiders in the various AI labs are suggesting that very intelligent AIs are coming very soon.

I wrote a bit about why this might be happening and what we can take away from their apparent confidence. www.oneusefulthing.org/p/prophecies...

About 15 minutes in, the lecture from @sendhil.bsky.social was a MasterClass on AI and how it will impact on economic research. www.aeaweb.org/webcasts/202...

Yeah, I understand. You could always 'approve' and send them a private message on the platform to pay more attention in future studies (or something along these lines). Attention-checks are also useful for this exact reason (provides a legitimate reason to reject).

I might be wrong, but rejections are actually a good way to weed out 'bad' Prolific users. After a certain number of rejections, they no longer get asked to participate in research. A good incentive in my view to provide high quality responses.

An interesting case of bracketing. Framing the cost 'narrowly' on each cigarette makes the cost more tangible and immediate for smokers.

Public policy can shift green behavior 👏

🚴 Promote cycling:
➡️ 1442km of bike lanes

🚶‍♀️Promote walking:
➡️Pedestrianized zones
➡️Speed limit reduction
➡️Low-emissions zones

🚗Discourage gas vehicles:
➡️Restricted traffic
➡️Phase out older diesel cars
➡️Removed 70,000 parking spaces
➡️Increased EV charging

The infected blood inquiry: Impact on public perceptions of blood supply risk, safety, and donation attitudes Background The UK's Infected Blood Inquiry (IBI) highlighted a major public health scandal, with at least 30 000 people infected and more than 3000 deaths attributable to infected blood and blood pr...

This study provides a preliminary gauge of public sentiment in a convenience sample. Further research is needed to assess the IBI’s long-term effects on perceptions & behaviours.

📄 Full article: doi.org/10.1111/tme....

5/5

Takeaways:

Despite the findings, UK participants continued to perceive blood supply safety as high & risk as low. Results may stem from:
- Public perception that the inquiry was thorough and fair.
- Modern safety standards vs. 1970–90s.
- Competing media (UK general election announcement)

4/5