[HN Gopher] Our brains create mental "chapters" with new event s...
       ___________________________________________________________________
        
       Our brains create mental "chapters" with new event segmentation
       study
        
       Author : domofutu
       Score  : 174 points
       Date   : 2024-12-15 05:29 UTC (4 days ago)
        
 (HTM) web link (www.psypost.org)
 (TXT) w3m dump (www.psypost.org)
        
       | domofutu wrote:
       | original study - https://news.ycombinator.com/item?id=42421683
        
       | aussieguy1234 wrote:
       | And now, AI LLMs have a way to prioritise what's important and
       | organise their memory (context window)
       | https://news.ycombinator.com/item?id=42411409
        
       | paweladamczuk wrote:
       | Please fix the title.
        
       | azeirah wrote:
       | I wonder if this is the same for people who haven't grown up with
       | linear media (books with chapters, shows with episodes/chapters,
       | albums with songs etc)
       | 
       | I always think of the WEIRD people nuance when reading about
       | these kinds of findings. Is the study cohort
       | 
       | - Western
       | 
       | - Educated
       | 
       | - Independent
       | 
       | - Rich
       | 
       | - Democratic
       | 
       | ?
        
         | HPsquared wrote:
         | As most of the readers are also WEIRD, the results would still
         | be relevant even if the scope was limited to that.
        
       | baxtr wrote:
       | So many upvotes but very few comments.
       | 
       | My feeling is that people are interested in the topic per se but
       | struggle with the takeaway from this paper.
       | 
       | I have tried to read the article and also skimmed the original
       | paper, but could not summarize any of it if you asked me now.
        
         | spacemanspiff01 wrote:
         | So what I got (but I don't really know what I am talking about)
         | 
         | When thinking about a story (or a sequence of events) we can
         | parse it in different ways.
         | 
         | We may think about the changes in time/location:
         | 
         | -------
         | 
         | 1 he went to the store
         | 
         | 2. then went to gas station
         | 
         | 3. then went home
         | 
         | 4. the next day he woke up.
         | 
         | ---------
         | 
         | Or you could think of the emotional changes:
         | 
         | ---------
         | 
         | 1. he was concerned about being unemployed.
         | 
         | 2. then at the gas station he got a call saying he got the job
         | and he starts tomorrow.
         | 
         | 3. He is nervous about his first day of work.
         | 
         | ------
         | 
         | With fmri we can detect these context shifts.
         | 
         | The thing the paper adds is that by prompting the user to pay
         | attention to 'emotions' or 'location' it affects where these
         | segmentation changes occur in the fmri results.
        
           | baxtr wrote:
           | Ok interesting, thanks.
           | 
           | In storytelling these are called outer and inner journey
           | respectively.
        
             | alexpetralia wrote:
             | Also called the fabula and syzuhet!
        
         | uxhacker wrote:
         | You're absolutely right--this topic is fascinating but complex,
         | which might explain why people are engaging with the idea but
         | struggling with the takeaway. One way to think about the paper
         | is that it extends the chunking model, which is essential not
         | only for UX design but also for understanding how people
         | perceive and organize the world around them.
         | 
         | How This Extends the Chunking Model
         | 
         | 1) Dynamic, Real-Time Chunking: Traditional chunking models
         | focus on discrete, often static information (like remembering a
         | phone number or words). This study shows how the brain applies
         | a similar principle to continuous, dynamic experiences--
         | segmenting life into "mental chapters" as events unfold.
         | 
         | 2)Neural Foundations: By identifying the specific brain areas
         | (hippocampus and prefrontal cortex) responsible for marking
         | event boundaries, it adds a biological layer to the chunking
         | model. It's not just a cognitive strategy--it's a neural
         | process tied to memory formation and retrieval.
         | 
         | 3) Application to UX Design: In UX, understanding event
         | boundaries can inform how we design user journeys. For example,
         | breaking a process (like signing up for a service) into clear
         | steps with defined "boundaries" helps users perceive the flow
         | and remember their progress. Similarly, designing interfaces to
         | reflect natural event segmentation (e.g., transitions between
         | scenes in a video or steps in an onboarding flow) aligns with
         | how people mentally organize experiences.
         | 
         | 4) Understanding and Communication:
         | 
         | Beyond UX, this model shows why clear, structured narratives
         | are critical for teaching, storytelling, or even summarizing a
         | paper. Without clear boundaries, information gets lost in the
         | noise. Potential Hypotheses for Altered States:
         | 
         | It also hints at why psychedelics, which might blur these event
         | boundaries, lead to a sense of timelessness or
         | interconnectedness. This could extend into therapeutic
         | applications or understanding atypical cognition.
         | 
         | Why This Matters for Understanding Things When we fail to
         | structure information with clear boundaries, it becomes harder
         | to process or remember--perhaps like your experience skimming
         | the article. This study offers a roadmap for how we can improve
         | communication and design to better align with our brain's
         | natural segmentation processes.
        
           | uxhacker wrote:
           | Really interesting is that current LLMs don't explicitly use
           | chunking for storage; they rely on distributed
           | representations across parameters. However, their self-
           | attention mechanisms and sequence processing mimic chunking
           | during runtime, creating dynamic "chunks" of context.
           | 
           | I'm at my limit, wondering if future models incorporating
           | explicit chunking for better memory, scalability, and
           | efficiency could truly take them to the next level.
        
           | flocciput wrote:
           | ChatGPT spotted.
        
             | baxtr wrote:
             | I had the same thought.
             | 
             | Funny how we humans have learned to spot AI generated
             | content!
        
         | fruit_snack wrote:
         | There's an interesting podcast episode of Lex Fridman with
         | Charan Ranganath (memory researcher) in case people are looking
         | for more on the topic
        
         | HarHarVeryFunny wrote:
         | The "graphical abstract" at the start of the linked article
         | seems to be the intended takeaway.
         | 
         | It seems to be a somewhat predictable consequence of the fact
         | that perception appears to work by us predicting what we're
         | seeing/experiencing, with sensory input merely
         | guiding/correcting those predictions. This is also related to
         | us mostly only seeing/remembering what we're focused on (i.e.
         | focused on predicting).
         | 
         | So, if we're predicting/focusing on an episodic experience
         | fitting some sequential "template" (e.g. restaurant or
         | proposal), then that's how we'll perceive it and
         | segment/memorize it.
        
       | m0llusk wrote:
       | This reminds me of the emerging description of conscious function
       | from Daniel Bor's book The Ravenous Brain. In this interpretation
       | the mind is constantly trying to interpret sensory data as intent
       | driven actions from differentiated actors. This makes available
       | critical information about threats and possible targets and so
       | on.
        
         | baxtr wrote:
         | Interesting.
         | 
         | Reminds me of the book "The Mind is Flat" by Nick Charter who
         | says that the mind is constantly improvising.
        
       | MailleQuiMaille wrote:
       | Well...if you present stories, for sure we gonna make chapters
       | out of it. Or beats, even.
       | 
       | I wonder what the results would have been if people were showed
       | documentary footage with no narration ? But my suspicion is that
       | just like we sometimes see human faces in places they clearly
       | don't belong, structuring information in a story format
       | (beginning, middle, end with rises and falls in between) is an
       | intrinsic part of how we process.
       | 
       | Maybe it's not so much that we like stories, but that we see
       | stories everywhere and the more information takes this digest
       | form, the more we feel at ease ?
        
         | HarHarVeryFunny wrote:
         | We always predict as well we can - that's just how the brain
         | works. Even without narration we'll be subconsciously comparing
         | what we're seeing to past experiences and using those
         | experiences to predict what comes next at various levels of
         | abstraction.
         | 
         | I highly doubt we have any intrinsic bias towards perceiving
         | things as following a story template, since our DNA has been
         | shaped by nature, not stories. The major bias we do have,
         | encoded in the way that our brain works, is just that nature is
         | largely predictable on various scales - next time will be the
         | same as last time - and this bias is what causes us to predict
         | and perceive/segment current experience based on past
         | experience.
        
           | MailleQuiMaille wrote:
           | >our DNA has been shaped by nature, not stories. Interesting
           | point, I believe the opposite. Or, let's say, that stories
           | have much more impact that DNA on our behaviour/thinking
           | models.
           | 
           | Religion, money, appartenance to a tribe outside of immediate
           | family, all of that are stories that we adhere to. Hell, look
           | at kamikazes : a group of people willingly destroying
           | themselves (and therefore, their DNA) for the perceived well-
           | being of a larger imaginary group, "their countrymen".
           | 
           | No, I believe animals can and do predict their environment,
           | but we differ because we can adhere to a layer of information
           | that is on top of what we can observe : call it collective
           | subconscious or myths, but this is information that helps us
           | do more.
        
       | tinthedev wrote:
       | This sounds quite interesting even from a non-AI/machine learning
       | perspective (as a lot of posts are discussing that).
       | 
       | I'd say it's a quite important lesson in UI/UX design, and
       | interfaces generally. As per the article, our brains are primed
       | to:
       | 
       | > actively constructs event boundaries based on internal
       | priorities rather than passively responding to environmental cues
       | 
       | Explains why ads effectivity is low, why a lot of users miss
       | obvious cues, and why UI design is so hard. It's something a lot
       | of professionals know intuitively and through anecdotal evidence.
       | 
       | It is quite interesting to see research dig deeper into it, and
       | possibly research/develop better "levers" to pull human attention
       | more efficiently - essentially make us remember things. Priming
       | was long understood to help, and that's primarily from
       | observation/empiricism. Deeper insights into how the brain itself
       | handles this could be game changers.
        
         | notnaut wrote:
         | Outside of a direct connection into the brain, doesn't it seem
         | a bit unlikely that we ever discover some sort of psychological
         | method of achieving significantly better attention grabbing for
         | individuals or (harder yet) groups than what we've already got?
         | 
         | The stuff my cartoon addled mind can picture is akin to
         | brainwashing with a black and white spinning spiral. That or
         | extensive, repeated exposure to certain stimuli, clockwork
         | orange style.
         | 
         | Better priming seems like the best we could hope to achieve, in
         | a sense, outside of things we'd probably be better off not
         | getting into?
        
           | jvanderbot wrote:
           | Not an expert, but it seems the Holy Grail would be to make
           | an advertisement relevant to the activity the person is
           | currently doing, or things they are thinking about. At least
           | then there's a better chance they'd remember it, if they in
           | fact remember the thing at all. Like a tool that recommends a
           | better tool when you try to do it for a job it's not intended
           | for. I always wish I had a better chop saw when I'm using a
           | circle saw for a too-large post, but when I see ads for chop
           | saws I ignore them.
           | 
           | This is probably why there's this enduring trope about going
           | to the hardware store and getting stuck for hours and walking
           | out with too much stuff. You're already in the project mode
           | and there to buy things you need. Very easy to grab just one
           | more thing. Like, why is hobby lobby still a thing?
        
           | DCH3416 wrote:
           | We have direct access to people's visual cortex and audio
           | processing with handsets. Folks are receiving a stream of
           | data tailored specifically to their life and experiences.
           | It's pretty direct while still being indirect.
           | 
           | A simple example in legacy media is with Coca Cola. The ads
           | show good experiences and attempt to anchor those emotions to
           | real life events, and then the tag line Enjoy. So your
           | enjoyment is tagged with having a Coke. Relatively straight
           | forward.
           | 
           | These days, and this is still an emerging technology. Ads can
           | be built and constructed on a per user basis. So rather than
           | generalizing, you can synthetically anchor ideas onto
           | individual real life emotions. And then at the right time
           | have the systems massage in the idea of compulsively making a
           | purchase. So while not strictly black and white spinning
           | spiral. It's more interception at a particularly vulnerable
           | moment. At least in my observations.
        
       | lenerdenator wrote:
       | I feel like my ADHD brain creates a bunch of sticky notes that
       | then get attached to a page in a notebook, which is then
       | unceremoniously shoved into a backpack with its cover open so
       | that the sticky notes come off of the pages and fall into the
       | backpack and get crumpled.
        
         | galleywest200 wrote:
         | It either does this, or it superglues the sticky note to the
         | zipper of the backpack so its always the thing you think about
         | first.
        
       | constantcrying wrote:
       | I hate these studies. The title is obviously false, the study
       | obviously could not demonstrate the claim the title makes.
       | 
       | Can we _please_ stop asking people something while looking at
       | their brain  "activities"? It is not helpful at all, it generates
       | exactly zero knowledge.
       | 
       | This is like trying to figure out how the weather works by
       | throwing scraps of paper into the air and tracing how the wind is
       | throwing them around.
        
         | jvanderbot wrote:
         | If you take everything literally then everything is wrong. All
         | models are wrong but some are useful. (And wasn't that
         | essentially the plot of twister?)
        
           | constantcrying wrote:
           | There is no model. It is just an observation that in March
           | more often then not the scraps thrown into the wind move in
           | an 8 like shape.
           | 
           | Even supposing this finding replicates, it shows essentially
           | nothing. It _definitely_ doesn 't show anything about the
           | brain creating "chapters".
        
       | giardini wrote:
       | IOW Minsky's frames and scripts, later fleshed out by Roger
       | Schank, but in an fMRI context.
       | 
       | https://en.wikipedia.org/wiki/Frame_(artificial_intelligence...
       | 
       | Before the recent LLM distraction, I thought that (frames,
       | scripts, et al; not fMRI) was the one true path.
        
       ___________________________________________________________________
       (page generated 2024-12-19 23:01 UTC)