[HN Gopher] How the brain navigates cities
___________________________________________________________________
How the brain navigates cities
Author : el_duderino
Score : 71 points
Date : 2021-10-19 11:41 UTC (11 hours ago)
(HTM) web link (news.mit.edu)
(TXT) w3m dump (news.mit.edu)
| mikewarot wrote:
| I had to traverse about 50 miles each way to a generating station
| for a year back in the 1980s. I learned how to get there from my
| boss, then quickly optimized the route as I drove it daily. After
| a week or two, I started searching widely for alternatives, and
| over time I just explored and picked the fun routes, or fast
| routes depending on my timeliness that day.
|
| I can reliably get back anywhere I've driven, assuming I started
| at home. On vacation it takes a day or two for that to be true. I
| doubt those now dependent on Navigation aids can do so.
| FabHK wrote:
| Tangentially related book recommendation:
|
| Jeff Hawkins, _A Thousand Brains: A New Theory of Intelligence_.
| It describes the importance of reference frames for intelligence,
| for locating limbs relative to the body, locating the body in
| space, locating objects in space (even abstract objects in
| abstract spaces), and a possible "implementation" of that in the
| columns of the neocortex (by, basically, if I understood
| correctly, positing certain neurons that encode location, eg via
| intersection of coordinates, and certain neurons that encode
| features, and then learning a connection between those).
|
| A short and interesting read (with quite some implications for
| progress in AI).
|
| Podcast with Sam Harris:
| https://samharris.org/podcasts/255-future-intelligence/
|
| Jeff Hawkins at Wikipedia:
| https://en.wikipedia.org/wiki/Jeff_Hawkins
|
| The tech developed by his AI company:
| https://en.wikipedia.org/wiki/Hierarchical_temporal_memory
|
| Book at Amazon: https://www.amazon.com/Thousand-Brains-New-
| Theory-Intelligen...
|
| ETA: Oh, quite some previous discussion on HN:
|
| https://news.ycombinator.com/item?id=20326396
|
| https://news.ycombinator.com/item?id=19311279
|
| https://news.ycombinator.com/item?id=26240901
|
| https://news.ycombinator.com/item?id=26794286
| codr7 wrote:
| It's all about finding the diagonals, I feel like a fool when I
| walk in squares for no good reason.
|
| Venice was very different though, there's water everywhere and
| you have to nail the bridges to get anywhere.
| ordu wrote:
| I remember while I was a kid, my father pointed me to a
| conundrum. In a front of our house there was a rectangle made of
| paths, and often we needed to cross this rectangle diagonally,
| but naturally we walked by its sides. And somehow we chose one
| path when coming home and another when moving from home. Why one
| path seems shorter when walking one direction and longer when
| walking the other one? I was grown enough to see that it is just
| plain stupid: both paths are of equal length. But at the same
| time one seemed shorter, and which one depended on a direction.
|
| He proposed a solution: we choose "more pointed" path, so we move
| a long side of rectangle first and a shorter one then. It is
| funny to see how scientists come to the same conclusion 30 years
| after my father.
| klodolph wrote:
| I remember going back and forth between the student center and
| the computer lab in college, in different buildings. I noticed
| that I took a different route in each direction... surely, not
| optimal.
|
| On my way to the student center, I'd leave the library by the
| nearest door. On my way from the student center, I'd enter the
| library by the nearest door. I figured that my brain was breaking
| it down into hierarchical steps: the computer center is inside
| the library, so the first step in getting to the computer center
| is entering the library, and I would do that by the nearest door,
| rather than choose a door close to the activities building.
| Brendinooo wrote:
| When I walk Manhattan (say, from Park and 42nd to 8th and 25th),
| I know that I need to cross a certain number of intersections to
| get there. It's efficient to cross wherever you see a walk sign
| early on, so most times you end up getting to 8th before you get
| to 25, because there are streets than avenues.
|
| So, "shortest path" might not be quickest path. And it might be
| an uglier or a more dangerous path; I think I've read elsewhere
| that people will often pick streets that look nicer or are more
| pedestrian-friendly.
|
| > Human beings are not optimal navigators.
|
| They should reconsider their definition of "optimal"!
| standardUser wrote:
| I used to walk across SF on a daily basis and I would usually
| go out of my way to avoid certain streets. Some because of
| homeless people/blocked sidewalks, others because of hyper-
| aggressive traffic. I'd happily add 10 minute to my walking
| commute just to start my day on slightly more pleasant streets.
| rob74 wrote:
| I often use a different "optimization": instead of crossing
| streets at intersections and waiting for traffic lights, I
| cross them mid-block when traffic allows. Of course that's
| easier to do in Munich than in New York...
| lexapro wrote:
| >Of course that's easier to do in Munich than in New York...
|
| That really depends on the street.
| TeMPOraL wrote:
| I do the same (Krakow). I also prefer to do "mid-block"
| crossing on speed bumps, because the traffic is guaranteed to
| be slow, and sometimes make an impromptu decision to cross
| the street earlier on a lights-free intersection when I spot
| a bus or a tram making a turn upstream, as it takes its time
| and blocks off incoming traffic for the duration.
|
| Humans have a complex goal function when optimizing their
| route, and they do online optimization (i.e. adjust as they
| go).
| micromacrofoot wrote:
| right, my route changes throughout the year as I optimize for
| more/less shadows to avoid hot/cold
| TeMPOraL wrote:
| I did that as a teenager, because I had stable routes
| (to/from school at roughly the same times). As an adult, I
| maintain awareness of the sun's position and keep
| following/avoiding the shadows on hot/cold days, sometimes
| crossing the street back and forth or in advance of a shade
| I'd like to take/avoid.
|
| It's almost an automatic process, but it has a visual
| perception component - on hot days, the lit side looks "too
| bright" to me, and on cold days, the shaded side looks "too
| dark" - so I take the side that looks "just right".
| tvanantwerp wrote:
| I'm always surprised that more people don't do this. It will
| be a hot sunny day, and I'll see people standing at an
| intersection waiting to cross. They'll be out in the sun,
| when mere feet away are shady spots they could wait in. I
| don't understand how/why they tolerate the heat instead of
| expending the tiniest effort to get in the shade.
| micromacrofoot wrote:
| it's pretty high on the hierarchy of needs, most people are
| just preoccupied... I tend to over analyze small
| optimizations like this to exert control over an otherwise
| chaotic reality
| nazz wrote:
| Some people want to tan; others are trying to avoid it.
| Arrath wrote:
| Personally, I have a pretty wide range of tolerance thanks
| to a career of working in the field so unless it is truly
| unbearably hot I just may not bother to find shade if I'll
| only be standing there a minute or two waiting for the
| light to change for me, for example.
| mackatsol wrote:
| My kids gave it a name: shade hopping
| 323 wrote:
| I once tried to move "forward" through old EU city streets, and
| instead managed to move in a circle and come back to the starting
| point, because I got confused about which way "forward" is, due
| to non-right angles between streets and lack of known to me
| landmarks.
| Arrath wrote:
| I once moved to a new city and immediately started a graveyard
| shift job. Since I was basically never out and about during
| daylight hours, for the first few weeks something was
| just...off. I couldn't gel my location and directions properly.
|
| Until I pulled up google maps and found, to my chagrin, that
| without having seen the sun to 'calibrate' myself, I had my
| cardinal directions off. Once I realize what was actually
| north, I was good to go.
| tobylane wrote:
| I wonder how this is affected by one way (for cars) streets. Do
| people think their way there by car then walk along those
| pavements?
| jgilias wrote:
| I find the whole premise that we're not trying to follow the
| shortest path weird. In our _evolutionary_ surroundings if you
| turn towards the thing you want to get to, and then walk, you're
| pretty much following the shortest path.
| asdff wrote:
| Maybe on a featureless frictionless plane that is true. There
| are hills, woods, rocks, rivers, all sorts of geographical
| barriers that make the easiest path often a longer one. Even
| with our technology today, mountain roads switchback rather
| than ascend directly to the summit.
| gnatman wrote:
| Related: Navigation-related structural change in the hippocampi
| of taxi drivers (2000)
|
| Structural MRIs of the brains of humans with extensive navigation
| experience, licensed London taxi drivers, were analyzed and
| compared with those of control subjects who did not drive taxis.
| The posterior hippocampi of taxi drivers were significantly
| larger relative to those of control subjects. A more anterior
| hippocampal region was larger in control subjects than in taxi
| drivers. Hippocampal volume correlated with the amount of time
| spent as a taxi driver (positively in the posterior and
| negatively in the anterior hippocampus). These data are in
| accordance with the idea that the posterior hippocampus stores a
| spatial representation of the environment and can expand
| regionally to accommodate elaboration of this representation in
| people with a high dependence on navigational skills. It seems
| that there is a capacity for local plastic change in the
| structure of the healthy adult human brain in response to
| environmental demands.
|
| https://www.pnas.org/content/97/8/4398
| mlac wrote:
| Having lived in Cambridge for a year, the shortest path is not
| always the most efficient path due to stoplights and traffic.
| It's a bit like UPS only making right turns.
|
| Also, there are streets that you would totally avoid if you need
| to stop in and pick something up, or get coffee (not sure on the
| fidelity of their data). Or to avoid walking down a basically
| empty road with no exits (just lined by buildings) vs. walking
| down a more populated road with store fronts.
|
| It's interesting, and humans are sub-optimal navigators on some
| dimensions, but not all of them.
| hinkley wrote:
| I recall as a young child, first learning about the Pythagorean
| theorem, and the moment I realized that cutting across a
| parking lot is almost always a 'short cut'. It was like I had
| discovered a secret of the universe.
|
| Quite different experience as an adult navigating a dense urban
| core and realizing that I could probably get to the Starbucks 4
| blocks up and 3 blocks over faster than I could get to another
| one that was a straight shot, even without shortcuts, just
| because of the damn crosswalks.
| brudgers wrote:
| This is evergreen. MIT has been singing this tune for more than
| sixty years. Since Kevin Lynch's _The Image of the City_ was
| published by MIT press in 1960.
| https://en.wikipedia.org/wiki/The_Image_of_the_City
|
| A great book (or as Bezos wants us to say, "a Goodread"). Highly
| recommended if you're interested in this sort of urban analysis.
| My architectural mentor, the late David Crane, FAIA was Lynch's
| research assistant while a grad student at Harvard;'s GSD. He's
| listed in the acknowledgements.
| SamBam wrote:
| This is only tangentially related, but I wonder if other people
| have noticed the phenomenological changes that happen as you
| become familiar with a city.
|
| When I come to a new city everything seems "flat," for want of a
| better word. The four directions at an intersection all have the
| same weight and pull and elevation.
|
| Once I get to know a city, my brain starts to introduce a sense
| of "upwardness" and "downwardness." Usually this is simply the
| cardinal directions, but not always, and in any case it rarely
| relates to actual changes in elevation. Once I know a city well,
| it's impossible to get the sense of "flatness" again, except for
| rare occasions when I come out of a subway stop and I'm
| disoriented, and then it can feel like a new city again.
| zemo wrote:
| yeah that's sorta what Kevin Lynch's book "The Image of the
| City" is about:
| https://en.wikipedia.org/wiki/The_Image_of_the_City
|
| very influential book
| dang wrote:
| Related:
|
| _Vector-based pedestrian navigation in cities_ -
| https://news.ycombinator.com/item?id=28915494 - Oct 2021 (6
| comments)
| jonbaer wrote:
| I feel like there are some flaws in this, we don't usually have a
| "top-down" look at a mile around us if not given information (ie.
| fog of war), feels like we tend to point A to point B repeat
| because we are comfortable to take a route until otherwise told
| by someone else there is a shorter path. For example if someone
| had a drone/scout around I am sure you would be able to determine
| best routes and chokepoints, traffic, etc.
| ARandumGuy wrote:
| I think this is a big factor. If I'm pointed towards my
| destination, then I always know I'm making progress, and it's
| easy to stay oriented. In a less "direct" (but shorter) route,
| it may not be obvious if you make an incorrect turn.
|
| I think another factor is just habit. The first few times you
| take a walk, you'll likely take the "easiest" route, even if
| it's not optimal by other measures. On subsequent trips, you'll
| likely just use the route you already know, instead of
| experimenting with different paths.
| padolsey wrote:
| I see this is about point-to-point navigation optimization but
| it's very curious this article doesn't mention grid cells -
| essentially, neurological ~1:~1 representations of physical
| space.
|
| > They were awarded the 2014 Nobel Prize in Physiology or
| Medicine together with John O'Keefe for their discoveries of
| cells that constitute a positioning system in the brain. The
| arrangement of spatial firing fields, all at equal distances from
| their neighbors, led to a hypothesis that these cells encode a
| neural representation of Euclidean space.[1] The discovery also
| suggested a mechanism for dynamic computation of self-position
| based on continuously updated information about position and
| direction.
|
| https://en.wikipedia.org/wiki/Grid_cell
| throwaway879080 wrote:
| there are also "vector cells"
| https://www.nature.com/articles/s41593-020-00761-w
|
| Neuronal vector coding in spatial cognition:
| https://www.nature.com/articles/s41583-020-0336-9?proof=t
| dendrite9 wrote:
| This article indicates a 1:1 relationship doesn't apply in 3D.
| I haven't had time to read and understand this enough but it
| seems worth passing on the article since its something on my
| list to dig into this weekend.
|
| "But when researchers were finally able to record from grid
| cells in animals navigating 3D spaces, the findings got "much
| more dramatic," Ulanovsky said -- seeming to demonstrate not
| just deviations from the framework, but departures from it. ...
| To their surprise, the hexagonal patterns that defined the
| cells' behavior in 2D were gone entirely: The researchers
| couldn't find even traces of that global order. Instead, the
| clumps of grid cell activity seemed to be distributed
| throughout the three-dimensional space at random. "Some
| properties were preserved," Jeffery said, "but the most
| visually striking property of grid cells was not.""
|
| https://www.quantamagazine.org/how-animals-map-3d-spaces-sur...
| blamazon wrote:
| Actual paper:
| https://www.nature.com/articles/s43588-021-00130-y.pdf
| com2kid wrote:
| The Seattle area (though not signifcant swaths of Seattle itself)
| are built on a cardinal grid, meaning you can get pretty much
| anywhere with just 1 turn, ignoring dead end streets.
|
| It is pretty nice, it means getting lost around here is _hard_.
|
| Our addressing system is absurdly logical, first 2-3 digits of
| house address are the street #, then you have the cross street as
| the 2nd part of the address. (Named streets actually are also
| numbered on the grid, and you can see the real street # above the
| name sign!)
|
| This means you can give me an address and I:
|
| 1. Know the exact location 2. Can navigate there
|
| Again with a sad exception for named streets, you'll have to look
| up the # of the street first before you head out.
|
| Because everything is on a cardinal grid, you just head
| north/south and then east/west.
|
| The best part is, if you get lost, it is easy to re-orientate
| yourself!
|
| If you get wet and the water tastes salty, you went too far west!
|
| If you get wet and it is fresh water with a hint of sewage, you
| went too far east!
|
| If everyone around you gets really polite, you are way too far
| north and have ended up in Canada, and if you see gas station
| attendants pumping gas, you've gone too far south and have landed
| in Oregon. Easy peasy!
| jugg1es wrote:
| There is a quote saying 'humans are not optimal navigators'...
| but it just depends on what you are optimizing for. We optimize
| for minimal cognitive load.
| kibwen wrote:
| _> Several years ago, the lab acquired a dataset of anonymized
| GPS signals from cell phones of pedestrians as they walked
| through Boston and Cambridge, Massachusetts, over a period of one
| year._
|
| I sure hope those people gave their consent and were adequately
| compensated for their data, because repeated observations as
| people walk between their home and their workplace is among the
| least-anonymizable data I can imagine.
| foo92691 wrote:
| I have news for you...
|
| https://techcrunch.com/2019/05/16/a-year-after-outcry-carrie...
| zemo wrote:
| while I like the topic and think this is an interesting lens,
| this study seems to start from the position that navigation can
| be understood to be a geometry problem alone, but uses sample
| data from the real world, which is not an abstract environment
| lacking in local qualities. This seems like a pretty large
| mismatch between the study's lens and the data set in question.
| I'm reading both this posting and the study here:
| https://www.nature.com/articles/s43588-021-00130-y.pdf
|
| > pedestrians appear to choose paths that seem to point most
| directly toward their destination, even if those routes end up
| being longer. They call this the "pointiest path."
|
| The sample in the article for how people walk versus the shortest
| path has a lot of other attributes beyond "it looks like it
| points there". The example trip is in Boston, the destination is
| probably the Boston Marriott on Huntington Ave. Who is taking
| this path? Have they spent time in the area, or is it their first
| time here? At what time of day are they walking? Are all path
| segments similarly well-lit? Do they have any sense of urgency?
| Are all path segments equally interesting?
|
| There is a reward for walking down an interesting street: it's
| interesting. You're stimulated. Is there a reward for making it
| to your destination in two minutes and ten seconds instead of two
| minutes and twenty seconds?
|
| In the example trip, pedestrians along the sub-optimal path are
| mostly taking Dartmouth Street, which is a main street for that
| neighborhood. Most people who have spent any amount of time there
| will know that Dartmouth hits Huntington, so you can just take
| Dartmouth; a local would give directions as such, and a local
| would probably walk as such. Their optimal route along Canton and
| Harcourt Street requires pedestrians to twice know that there is
| a path that can continue when it does not appear on a street map;
| twice the path is disjoint. This doesn't appear to be
| acknowledged in the study. You would have to know the side
| streets very well to know that path was even an option, and
| explaining it to a tourist would be very cumbersome and involve a
| lot of steps. For example, assuming the pedestrian is exiting a
| building facing south, the path chosen by pedestrians can be
| explained thusly: "make a right when you get outside, then take
| your first right along Dartmouth, and keep going until you see a
| big intersection and make a left on Stuart". The optimal route:
| "make a right when you get outside, when the street ends take a
| right onto Cantor, and then stay on Cantor until it ends, cross
| over the pedestrian zone and you'll be on Harcourt. When you get
| to the corner of Harcourt and St Botolph, go over the sidewalk to
| continue along Harcourt because Harcourt is disjoint. Keep going
| along Harcourt until you reach the six-way intersection, and make
| a slight right onto Huntington." I know how ridiculous this is
| because I lived over there for five years. To compare these two
| routes as geometric shapes and ignore their actual features is of
| unclear value. The route preferred by Google Maps to go from 79
| Warren Ave to Boston Marriott Copley Place is the shortest route;
| the route that people do _not_ take.
|
| > pedestrians chose routes that were slightly longer but
| minimized their angular deviation from the destination
|
| this is one way of looking at it, but another way of looking at
| it is that people take main roads because they know how the main
| roads connect but aren't sure which sidestreets connect and which
| ones have dead ends. The study also doesn't seem to mention time
| of day for the data and density of street lights; a more direct
| route going through an unlit street is likely not going to be the
| preference of a person walking at night.
|
| Anyway, there is a very good book called "The Image of the City"
| by Kevin Lynch, which is about how people form mental models of
| the built environment. People use these mental models to
| navigate. One of the model cities in that book is also Boston,
| and the book also came from MIT. It is curious that none of the
| concepts mentioned in Lynch's book appear in this study; it seems
| to suggest that the authors believe that people only imagine the
| geometry of a space.
| https://en.wikipedia.org/wiki/The_Image_of_the_City
| WouterSpaak wrote:
| I wonder if there's also a preference in the mind to stay on a
| particular street as long as possible, because the street
| architecture, street furniture and amount of traffic stays
| roughly the same, and therefore predictable and easy to walk on.
| w-m wrote:
| Like the example given in the article, I also take different
| routes from and to work regularly, with my bicycle. This habit
| developed naturally, I didn't really plan it. I was reflecting
| on why I do this, and your explanation of street continuation
| seems to be a better fit.
|
| When I start in either direction, I do not have a clear heading
| of the destination, as there is no line of sight, so the
| pointiness from the article makes less sense. I guess
| continuing forward on the same street feels getting there
| faster, as making a turn feels slow.
___________________________________________________________________
(page generated 2021-10-19 23:01 UTC)