Louis Tay

Your brain will choose self-maintenance over your performance. Every time. Sleep isn't a luxury - it's the basis for functioning. wsj.com/health/wellnes…

A reminder that peak functioning also requires sleep. MIT researchers found that when you skip sleep, your brain forces emergency "cleaning cycles" during the day. It works to flush out waste that normally clears overnight. Each cycle briefly shuts down your attention.

3/ Without this clarity, scales look psychometrically sound but remain theoretically underdetermined. We measure something - but what does score variation represent? This also determines scale anchors and whether reverse-worded items even belong in your measure. #psychometrics

2/ What do the endpoints represent? Job Satisfaction: Is low = neutral or dissatisfied? Extraversion: Is introversion absence or opposite? Burnout: Does it run to "no burnout" or to "engagement"? Trust: Is distrust low trust - or distinct?

1/ We call constructs "continuous" but rarely specify what that continuity actually means. A neglected step in construct validation. 🧵 journals.sagepub.com/doi/full/10.11…

But active collaboration, where we interact with outputs and refine them, appears to be okay. Interestingly, passive use initially boosted enjoyment and satisfaction - but reversed when people had to go back to manual work without AI! nature.com/articles/s4159…

Do you use AI in your work? New research points to the importance of HOW we use AI. Passive use, where we just cut and paste outputs, reduces self-efficacy, psychological ownership, and work meaningfulness.

Or maybe part of it is something deeper: a normative intuition that some domains of human life shouldn't be directed by algorithms, regardless of how "good" AI gets. pewresearch.org/science/2025/0…

What type of AI use is more acceptable? Forecasting the weather is the most acceptable. But not advising people about their faith in God. Likely, part of it is about perceived capability: AI can model atmospheric data, but can it grasp the divine?

As psychologists, we should be shaping what "evidence-based methods" for AI crisis detection actually look like. apaservices.org/practice/busin… #AIPolicy #MentalHealth #Psychology

CA SB 243 now requires companion chatbots to implement crisis protocols for suicidal ideation, disclose AI identity to minors, and give families a private right of action.

New from APA: California, New York, and Utah are leading a wave of state legislation regulating AI chatbots and mental health.

I have personally experienced reframing many work and life demands as opportunities and privileges. While it hasn’t reduced the demands, it has changed how they feel and how I talk about them. #reframe #mindset

A “get to” mindset (more autonomous, approach-focused) frames those same demands as chosen or meaningful pursuits. It doesn’t eliminate stress, but it can change how conflict is experienced and narrated.

Sometimes the difference between thriving vs. not isn’t what we’re doing - but how we’re thinking about it and talking about it. nature.com/articles/s4159… In our Nature Scientific Reports paper, we examined how language on social media reflects work–personal conflict.

So, the data tells us lonely people find something in AI companions. It doesn't tell us they come out less lonely in the long run. sciencedirect.com/science/articl…

But here's the critical caveat - at least from what I'm reading: we don't actually know if AI use actually reduces loneliness, or if lonely people simply turn to AI more.

"What about people who have no one to talk to? Is an AI companion better than nothing?" This was a question posed to me after my talk.