[HN Gopher] The "computers are social actors" theory no longer a...
___________________________________________________________________
The "computers are social actors" theory no longer applies to
desktop computers
Author : vasco
Score : 52 points
Date : 2023-11-14 07:44 UTC (1 days ago)
(HTM) web link (www.nature.com)
(TXT) w3m dump (www.nature.com)
| cobaltoxide wrote:
| Abstract
|
| The Computers Are Social Actors (CASA) theory is the most
| important theoretical contribution that has shaped the field of
| human-computer interaction. The theory states that humans
| interact with computers as if they are human, and is the
| cornerstone on which all social human-machine communication
| (e.g., chatbots, robots, virtual agents) are designed. However,
| the theory itself dates back to the early 1990s, and, since then,
| technology and its place in society has evolved and changed
| drastically. Here we show, via a direct replication of the
| original study, that participants no longer interact with desktop
| computers as if they are human. This suggests that the CASA
| Theory may only work for emergent technology, an important
| concept that needs to be taken into account when designing and
| researching human-computer interaction.
| vlovich123 wrote:
| Was the original paper ever replicated? Cause if not, that's
| too strong a conclusion. A simpler explanation could be that
| the original paper got something wrong and wasn't replicable.
| weinzierl wrote:
| _" This suggests that the CASA Theory may only work for
| emergent technology[..]"_
|
| I think we can see this unfold in real time with large language
| model based chatbots like ChatGPT. First they seem almost human
| and the initial reaction is to treat them like that. Always say
| "Thank you!" and potentially get angry at them for being wrong.
| It doesn't take long though, to realize that the bot is a bot
| and even if it speaks human language it behaves significantly
| different from a real human. Then the human behavior starts to
| change as well and the bot is treated differently.
| pixl97 wrote:
| Except we see to see behaviors in asking questions to LLMs
| where we get different/better responses if we ask "please".
| chihuahua wrote:
| And also, we get better answers from LLMs for solving
| captchas if we claim that the hard-to-decipher letters were
| written on our dear grandmother's Christmas ornament. I
| find this quite amusing.
| l33t7332273 wrote:
| This makes sense if you think about chatGPT as a ML model
| and not as a sentient AI. Its training data would show that
| asking nicely elicits better answers.
| kmeisthax wrote:
| >One main reason why no direct replication has been conducted
| before now, is that the proposed underlying psychological basis
| for the CASA Theory suggested that a short period of time (e.g.,
| 30 years) should have no influence on the effect. Specifically,
| the original authors postulated that the reason we respond to
| social computers as if they have an awareness is because our
| brains are not evolving at the same rate as technology; our
| brains are still adapted to our early ancestors' environment4.
|
| I find this hypothesis... shockingly naive, bordering on
| psuedoscientific. Brains are there _to learn behaviors faster
| than evolution can_. Neuroplasticity (as well as just the general
| experience of _being human_ ) means that brains can un- and re-
| learn things as needed rather than being stuck with the same
| static behaviors that, say, a plant might have. No surprise the
| CASA theory got falsified, and good on the researchers for doing
| the necessary falsification work.
| amelius wrote:
| Perhaps they were talking about test subjects that haven't yet
| been exposed to computers.
| mrdatawolf wrote:
| "Nass and Reeves make a point of stating in their methodology
| that all the participants 'have extensive experience with
| computers ... they were all daily users, and many even did
| their own programming'"
| KMnO4 wrote:
| My goodness, how did that pass peer review?
|
| Postulating that evolution is strictly the result natural
| selection (ie our ancestral chain is solely responsible for
| what we are capable of doing) is such an archaic view.
|
| We don't have to "randomly mutate and wait for the inferior
| genes to drop out of the gene pool" to see human adaption, and
| there are a plethora of counterexamples to remove any doubt.
| patcon wrote:
| I say this with utmost respect (and a little bit of devil's
| advocacy heh :) -- I might suggest your critique is a little
| too self-assured.
|
| Disclaimer: Not a neuroscientist, but a biochemist *shrug*
|
| We recognize faces easily due to deep wiring. Some linguistic
| behavior is deeply rooted as well, e.g., our ability to be
| immersed in written or visual narratives of experiences that
| are not our own ("aesthetic illusion") is an evolved extension
| of "play". Both of these are related to deep brain structures,
| not just culture. (Though HOW we play is certainly also in the
| realm of culture.)
|
| Lots of conclusions about brains can be rooted in there being
| really foundational neural structures that can't simply be
| rewired by culture. Just because culture exists, it doesn't
| mean that conclusions rooted in biology are irrational. And
| assuming the effect was reproduced over many cultures (which
| perhaps it was not) then it would be fair to assume it was
| rooted in some deeper system.
|
| tldr- I feel you are falling for hindsight bias. Respectfully,
| I suggest we're all best-served to navigate this world by
| cultivating a healthy dose of humility, and I say that more for
| all the self-assured readers upvoting you for making a dunk.
|
| I'm guessing the truth is somewhere in between :)
| acomar wrote:
| except no such neural structure has ever been found. humans
| have been using tools for longer than we've been human --
| without solid evidence that _this_ tool is interpreted as a
| social actor, based on real neuroscience, this kind of claim
| rooted in an evolutionary argument _is_ psuedoscience. people
| have been making arguments from evolution to say all kinds of
| nonsense things since Darwin (like justifying racial
| hierachies). which neural structure is posited to cause us to
| humanize our tools?
|
| if anything, the historical evidence points in the opposite
| direction -- that people objectify far more than they
| humanize, even when the cost is measured in hundreds,
| thousands, or millions of lives. that's merely an
| observation, not a hypothesis or a claim about what people
| will do or about what they are capable of doing. we ought to
| humanize more often.
| danaris wrote:
| "We haven't yet found a specific neural structure for
| recognizing faces" is far from evidence that no such
| structure exists. Our understanding of the brain still has
| massive gaps in it, and I can testify from my own
| experience working for a psychology & neuroscience
| department (which includes one person particularly
| specializing in perception, from the very basic "light
| hitting the optic nerve" stage all the way to object
| categorization and recognition) that we still have a lot to
| learn in this area specifically.
|
| It may very well be that there _isn 't_ a brain structure
| dedicated to this, and that would be fascinating, too! But
| to denigrate the people doing their best to understand this
| stuff 30 years ago as "pseudoscientific" just because they
| made an assumption about how plastic the brain was without
| our benefit of 20/20 hindsight is very much uncalled-for.
| acomar wrote:
| > "We haven't yet found a specific neural structure for
| recognizing faces" is far from evidence that no such
| structure exists.
|
| proving a negative is, famously, quite hard. an unsolved
| problem, even. facial recognition has a plethora of
| evidence beyond an argument from evolution. the notion
| that we humanize tools is one that, as yet, lacks that
| evidence. I urge people to be more skeptical of arguments
| from evolution. we understand very little about our
| evolution and it's easy to insert our own worldviews and
| beliefs into such arguments, allowing them to state
| virtually anything we like in a plausible envelope with
| the shape of a scientific argument. I'm not just calling
| the argument about humanizing tools pseudoscience -- I'm
| applying it equally to every other argument from
| evolution that lacks other motivating evidence.
| anonymouskimmer wrote:
| I understand your original point was about a neural
| structure involved in humanization, and not of facial
| recognition, but am responding to the point you let the
| interlocutor derail this to.
|
| > > "We haven't yet found a specific neural structure for
| recognizing faces" is far from evidence that no such
| structure exists.
|
| > proving a negative is, famously, quite hard.
|
| Whether structure or not, we do have very strong evidence
| that a _mechanism_ of facial recognition exists as there
| are people who lack this mechanism to various degrees.
|
| This article posits that we have indeed discovered a
| specific neural structure involved in facial recognition:
| https://www.aipc.net.au/articles/the-neuroscience-of-
| facial-...
|
| > The brain has even evolved a dedicated area in the
| neural landscape, the fusiform face area or FFA
| (Kanwisher et al, 1997), to specialise in facial
| recognition. This is part of a complex visual system that
| can determine a surprising number of things about another
| person.
| Legend2440 wrote:
| Your brain is a structure for learning structures.
|
| It doesn't need to have a built-in module for recognizing
| faces; it wires up a face-recognition system on the fly,
| from visual data.
| wolverine876 wrote:
| > humans have been using tools for longer than we've been
| human
|
| It depends on how you define 'human'. Our line spit from
| chimpanzees 7 million years ago (mya); we walked upright 6
| mya. Tool use began ~2.58 mya (possibly 3.3 mya, depending
| on some uncertain evidence).
| anonymouskimmer wrote:
| https://neurosciencenews.com/empathy-human-robots-
| psychology...
|
| > They performed electroencephalography (EEG) in 15 healthy
| adults who were observing pictures of either a human or
| robotic hand in painful or non-painful situations, such as
| a finger being cut by a knife. _Event-related brain
| potentials for empathy toward humanoid robots in perceived
| pain were similar to those for empathy toward humans in
| pain. However, the beginning of the top-down process of
| empathy was weaker in empathy toward robots than toward
| humans._
|
| So basically it seems we potentiate empathy toward similar
| kinds of beings and then maybe pattern-recognize that they
| are not similar to clamp down on the potentiated empathetic
| response?
| anonymouskimmer wrote:
| > Brains are there to learn behaviors faster than evolution
| can.
|
| Reflexes whether innate, enforced, or purely learned are
| basically impossible to modify without either conscious
| attention or a lot of training. I presume the CASA proponents
| were arguing that these social responses are akin to social
| reflexes.
|
| Plants "learn" too. https://theconversation.com/pavlovs-plants-
| new-study-shows-p...
|
| The mechanistic hypothesis being:
|
| > Plants may lack brains and neural tissues but they do possess
| a sophisticated calcium-based signaling network in their cells
| similar to animals' memory processes.
| fredgrott wrote:
| It is not just that as plant cell signaling is how animals
| got neurotransmitters through evolution. In fact many plants
| have neurotransmitter analogs due their signaling needs
| through evolution.
| layer8 wrote:
| Sounds more like this is just an exhibit of the replication
| crisis. I don't remember people interacting with computers as if
| they are human in the 1990s -- at least the people who actually
| did interact with computers on a regular basis.
| pixl97 wrote:
| I mean, it might have been before your time, but people have
| been doing this since the 60's.
|
| https://en.wikipedia.org/wiki/ELIZA
| layer8 wrote:
| I'm familiar with this of course, but it's conditioned on a
| chat program that actually tries to appear human, and the
| credulous reactions to ELIZA were mostly lay people who
| otherwise hadn't used computers before. Likewise, some people
| -- famously including some software engineers -- believe that
| certain AI chatbots are sentient, which also doesn't falsify
| the linked study. (This argument holds regardless of whether
| you consider AI chatbots to be sentient or not.)
| crote wrote:
| But what does that actually _show_?
|
| Humans have been communicating with other humans via text for
| _ages_. Fooling a human that the person on the other side of
| text communication is also a human isn 't too hard,
| especially when they are not looking for signs of deception.
|
| Even if they are aware it is a computer, humans will
| anthropomorphize literally anything. Animals are obvious, but
| we'll even extent it to simple objects like cars. We'll talk
| to them with zero expectation of any response or sign of
| sentience. It's a suspension of disbelief, really. It's not
| any different than kids playing Cops And Robbers, people
| performing a theatre play, or playing a board game with
| friends.
|
| CASA seems to suggest that humans will view _any_ computer
| interaction as one between two humans. That 's not just
| interacting with chat bots, that's also something like
| _opening the Configuration Panel_. If you ask me, that 's a
| massive stretch!
| Probiotic6081 wrote:
| Did you read the article? Did you understand it? CASA doesn't
| postulate that users in the 90s interacted with computers as if
| they were literally other humans.
| snakeyjake wrote:
| layer8:
|
| I don't remember people interacting with computers as if they
| are human in the 1990s -- at least the people who actually
| did interact with computers on a regular basis.
|
| You:
|
| CASA doesn't postulate that users in the 90s interacted with
| computers as if they were literally other humans.
|
| The article:
|
| The Computers Are Social Actors (CASA) theory is the most
| important theoretical contribution that has shaped the field
| of human-computer interaction. The theory states that humans
| interact with computers as if they are human, and is the
| cornerstone on which all social human-machine communication
| (e.g., chatbots, robots, virtual agents) are designed.
| However, the theory itself dates back to the early 1990s,
| and, since then, technology and its place in society has
| evolved and changed drastically.
|
| Me:
|
| Confused.
| epcoa wrote:
| From the original article: "The [CASA] theory states that
| humans interact with computers as if they are human, and is
| the cornerstone on which all social human-machine
| communication (e.g., chatbots, robots, virtual agents) are
| designed. "
| make3 wrote:
| I agree. this theory sounds ludicrous to me except for people
| who maybe have absolutely zero technical know-how
| wolverine876 wrote:
| That describes people using computers for the first time, a
| widespread experience when the first experiment was done.
| mrdatawolf wrote:
| True, but they did say the experiment was done with people
| who had been using computers regularly.
| wolverine876 wrote:
| Good point. Though they still could be pretty new - not
| like people who grew up with them.
| TeMPOraL wrote:
| That doesn't make sense. Computers aren't, and weren't even
| back then, some kind of alien technology, unlike anything
| else people have experienced earlier. I can't imagine those
| people treating their calculators, vacuums or cars as
| persons. A computer is a qualitative jump, but _not that
| big_.
| wolverine876 wrote:
| What is that based on? And what did they use that was
| like a computer?
| wolverine876 wrote:
| > Sounds more like this is just an exhibit of the replication
| crisis.
|
| Do you have evidence of that, or are you saying it looks the
| same?
|
| The replication crisis on HN is replication of these comments
| in every discussion of social science research. The papers that
| they tried to replicate at least were at least based on
| evidence, and the conclusion was not that they were completely
| wrong but that, for many papers (not all by any measure), the
| conclusion was replicated but the evidence wasn't quite as
| strong as in the original.
| layer8 wrote:
| There are two explanations for the discrepancy between the
| original study and the present one:
|
| 1) The behavior of people has changed substantively since the
| first study. (Which is what the present study seems to be
| concluding.)
|
| 2) At least one of the studies wasn't representative and/or
| didn't actually measure what it claimed to measure. (That is,
| an exhibit of the replication crisis.)
|
| What I'm saying is that explanation 2 seems much more
| plausible to me, given that the results of the first study
| don't match my recollection and experience at all.
| wolverine876 wrote:
| There are other explanations too, and every study flaw, if
| it exists, doesn't implicate the "replication crisis".
|
| > the results of the first study don't match my
| recollection and experience at all.
|
| 30 years ago you had some anecdotal impressions. 30 years
| ago the researchers had a hypothesis, focused on this exact
| question, designed an experiment, and collected evidence.
| (And many since have found it valid.) 30 years later, we
| have a detailed written report of their experience, and
| your 30-year old memory as posted on HN.
|
| It's just too easy to say, 'that's not my experience'
| without evidence. Everything is 'debunked' on that basis.
| Now we have 30 year old memories too.
| mvdtnz wrote:
| You know the replication crisis is a real thing, right?
| karaterobot wrote:
| It seems like they found that the CASA effect was real _when
| the technology was novel_. Once you get comfortable with
| computers, it wears off.
| ordu wrote:
| _> Sounds more like this is just an exhibit of the replication
| crisis_
|
| Yeah. In the article we can find that _" the original study
| recruited 30 Stanford University undergraduates (10 in each of
| the original 3 conditions, see Procedure)."_
|
| I wouldn't think twice before throwing away the results and
| forgetting about them.
|
| The replication study does better, recruiting 132 participants.
| This is much harder to ignore.
| svachalek wrote:
| From what I can understand, the original finding is along the
| lines of when the computer says "insert floppy disc 3", people
| instinctively responded as if it was a person giving them that
| order and subconsciously went through all the usual social
| processing for that response - do they have authority over me?
| is anyone else watching? etc. Not that they thought the
| computer actually was capable of human interaction at a
| conscious level.
|
| This is more believable than the way most people are
| interpreting it but I still have doubts. I was an adult in that
| era, and what I remember is computers were just so much more
| mechanical then. It was more like operating a microwave oven
| than today's much more sophisticated machines. It's hard to
| imagine those simple error codes triggering human social
| reactions.
| layer8 wrote:
| True. Though I would argue that people nowadays are generally
| less submissive to authority than 30 years ago, hence that
| might not be specific to computers.
| anonymouskimmer wrote:
| An alternate hypothesis is that CASA is still correct, but that
| people start interacting with computers in a less than
| compassionate manner. E.g. people may start to see themselves as
| "wealthy" compared to the computer (or whatever the general
| mechanism of a reduction in empathy):
| https://greatergood.berkeley.edu/article/item/how_money_chan...
| rossdavidh wrote:
| It's still a social actor, but I'm higher ranking, it's a
| peasant? Interesting, but hard to test.
| anonymouskimmer wrote:
| Maybe in the future for computers that are expected to make
| decisions directly affecting the welfare of humans.
|
| I think we would _start_ to see this in people defending the
| decision of an AI in arguments with other people (such as for
| automated vehicles). What it ultimately turns out to be I don
| 't know.
| dj_mc_merlin wrote:
| Seems like possibly bad timing giving advances in AI.
| ChatGPT/GitHub Copilot can already be used to interface with a
| computer (there's some git repos floating about if one looks), I
| think it's only a matter of time before someone makes an actually
| useful AI interface.
| make3 wrote:
| tool use is already that.
| danaris wrote:
| If you think it seems like bad timing, I suggest you read
| closer: They very specifically say that _emerging_ technologies
| (like the recent advances in ML systems) _are_ still subject to
| the CASA theory.
| rossdavidh wrote:
| My 18yo daughter, not tech illiterate but not particularly
| interested in programming, has mentioned to me that she found
| AI-generated artwork really interesting for about a month,
| but it quickly became boring. I wonder if it comes from a
| "wow what kind of creative person could come up with
| something that unusual?" initial response, which quickly
| gives way to boredom as one realizes that there is no
| creative person behind there?
|
| Also, I wonder if the period before "emerging" technologies
| stop seeming interesting (e.g. like another person) is
| speeding up over time, like so much else?
| layer8 wrote:
| I suspect it's more because most AI-generated artwork looks
| same-ish after a while. At least that's my own perception.
| rezmason wrote:
| In 2006 I shipped my iBook to Apple to fix its screen bezel. I
| made a Flash animation where my laptop cheekily described its
| issue, thanking the technicians in advance, and I made it launch
| on startup.
|
| The iBook came back with a factory reset. I restored from backup
| and stopped anthropomorphizing my computers.
| russellbeattie wrote:
| Humans seem to naturally anthropomorphize anything which exhibits
| systematic yet unpredictable or capricious behavior: Pets,
| machines, mountains, oceans, weather, planets, ex-wives, etc. We
| give them names and project a sense of agency to them, because
| they must obviously act certain ways some times on purpose. It's
| the reason we have gods, fairies and superstitions. It's why we
| name our cars and our storms. It's just what we do.
|
| So to me, it makes sense that as we understand something more,
| being able to predict what and why it does something, the less
| likely we are to think of it as having some sort of inner being.
| anonymouskimmer wrote:
| > So to me, it makes sense that as we understand something
| more, being able to predict what and why it does something, the
| less likely we are to think of it as having some sort of inner
| being.
|
| Pretty much the opposite with respect to pets and humans we are
| acquainted or intimate with.
|
| The sheer speed with which computers change (the modern
| automated update cycles) might be a counter to any impulse we
| have to anthropomorphize them. Even with animals we are much
| more likely to anthropomorphize longer-living animals than
| shorter living ones. This could even be something like a
| survival mechanism - don't become emotionally attached to
| something with you only a brief while.
| mvdtnz wrote:
| I've never heard this theory but it just sounds incredibly stupid
| and unrealistic to me. The kind of thing some academic nonce who
| never stepped foot off a university campus would come up with. I
| don't treat any computer I've ever owned like it was a "social
| actor". Computers are tools.
___________________________________________________________________
(page generated 2023-11-15 23:02 UTC)