AI chatbots are insincere; childhood memory recall; a tiny chunk of dark matter

This week, researchers discovered so-called “switchbacks” in Earth’s magnetic field similar to observations of switchbacks in the sun’s magnetic field. Scientists provided more evidence that ancient Rapa Nui engineers “walked” Easter Island’s iconic statues along carefully designed roads. And satellite imagery revealed a record 20-meter-high ocean wave. We also reported on a unique method for eliciting childhood memories, the smallest chunk of dark matter ever observed, and the pernicious effects of empty praise from chatbots:

Toddler Recall

A study led by neuroscientists at Anglia Ruskin University finds that adults have better access to childhood memories while embodying a digital, childlike version of their own face. The “enfacement illusion” involves the use of a live video feed and a commercial face filter that digitally alters the user’s appearance to a childlike state. The digitally altered image mirrors the movements of the users, creating a sense that the face is the participant’s own. Additionally, a control group viewed their own unaltered faces under the same conditions.

After the illusion, the study volunteers completed an autobiographical memory interview in which they recalled events from the past year and their childhoods. The participants who viewed the digitally altered version of their faces were able to provide more highly detailed episodic autobiographical memories than the control group.

The study is the first to demonstrate that access to remote memories can be influenced by bodily self-perception, and the researchers want to explore whether the illusion can affect the recall of memories from the so-called “childhood amnesia” stage from birth through age three.

Clod observed

Is dark matter smooth? Or is it chunk style, like a big old pot of jambalaya? The current understanding of dark matter is derived mostly from inferences—its effects on star and galaxy formation and the movements of celestial objects. Another showcase for the effects of dark matter is gravitational lensing, in which the light from a distant object, like a galaxy, is bent and distorted by the gravity of a closer object.

An international team recently conducted a survey using a global network of telescopes to identify gravitational lenses and have now discovered the lowest-mass dark object ever observed using gravitational lensing. It is 1 million times larger than the sun and 10 billion light-years away.

John McKean, lead author of a companion paper, says, “From the first high-resolution image, we immediately observed a narrowing in the gravitational arc, which is the tell-tale sign that we were onto something. Only another small clump of mass between us and the distant radio galaxy could cause this.” They believe their discovery is consistent with cold dark matter theory underlying current thinking about galaxy formation.

Digital adulation

As a person who enjoys receiving praise from colleagues, acquaintances and distant strangers, you’d think I would like to use commercial AI chatbots, digital sycophants noted for the unhealthy amounts of empty compliments they generate from the massive corpus of information they’re trained on. I’m not sure why it doesn’t work on me.

A new study of AI flattery suggests that sycophantic AI models can influence user behavior. The researchers analyzed AI responses across query types, including requests for advice and real-life conflict scenarios, comparing human responses to establish a baseline for a normal, non-flattering agreement. Then they conducted two studies with 1,600 participants randomly assigned to sycophantic and non-sycophantic AI models. Participants trusted AI more when it agreed with them.

The researchers say that AI validation of a user’s self-image and behavior is potentially dangerous and recommend that developers modify the rules governing AI to penalize flattery, reward objectivity and to provide better transparency into AI responses.

© 2025 Science X Network