enhance 15 to 23
nikola[at]klaxzy.net:/var/notes/voight-kampff-neo/
_
Voight-Kampff-Neo
"a very advanced form of lie detector that measures contractions of the iris muscle and the presence of invisible airborne particles emitted from the body. the bellows ere designed for the latter function and give the machine the menacing air of a sinister insect. the VK is used primarily by Blade Runners to determine if a suspect is truly human by measuring the degree of his empathic response through carefully worded questions and statements.[0]"
the test originally appears in Philip K. Dick's book Do Androids Dream of Electric Sheep?, where it is spelled "Voigt-Kampff."
the idea of using a voight-kampff-like polygraphs to identify AI online seems intriguing to me. in an era where deepfakes and ai generated content are increasingly prevalent, there is a growing need for reliable methods to distinguish between human and artificial entities. the VK polygraph, as depicted in the story, appears to offer a potential solution.
however, a closer examination of the concept raises several philosophical and critical concerns.
to be human
image: polygraph from the bladerunner movie used in the scene to perform a voight-kampff test
the Voight-Kampff test (VK in short) is based on a set of assumptions about how humans behave and react emotionally. but people from different cultures and backgrounds may exhibit different emotional responses. it's hard to define what makes someone human in a way that works for everyone. being human is a complex mix of cognitive, emotional, social, and moral aspects. a polygraph test can't really capture the full range of human experiences.
the test relies on a persons emotional response to a series of questions. but can a machine truly experience emotions like humans do, or is it just simulating emotional responses to mimic human behavior? what about people who are emotionally numb or exhibit a different emotional profile due to cultural or personal reasons? The test limitations in this regard are significant.
A final question, Two-part. You are watching an old movie on TV, a movie from before the war. It shows a banquet in progress; the guests are enjoying raw oysters. The entree, consists of boiled dog, stuffed with rice.
the line is from the scene where Bladerunner Rick Deckard perform te test to Rachel. the scene can be seen as a commentary on the human condition, particularly in relation to cultural norms and traditions. the fact that the banquet is serving boiled dog, which is typically considered a companion animal, however the dog meet is boiled as the meet should be usually, on the other hand it says that the guests are enjoying the raw oysters, here we have a raw meet instead. both options can be considered acceptable or disgusting for some, where some might say that eating a raw oyster is a very enjoyable thing but eating the dogs meat, in any form, is a disgusting immoral thing to do, however dog meet could be considered a specialty in some parts of south-east Asia. if you are vegan you would probably reject both.
here, the tests relies on cultural norms and assumptions, which can vary greatly across different societies and contexts. a human raised in a culture where eating dogs is common would likely not be shocked by the boiled dog scene.
from a psychological perspective, the scene can be seen as a reflection of the themes of identity, empathy, and compassion, this is introducing plenty of emotional complexity, humans are capable of complex emotional responses, which cannot be reduced to simple yes/no answers. humans can be shocked, offended, or disgusted by the boiled dog scene, but can also empathize with the replicants predicament.
image: photographic hoax showing the announced opening of a Great Dog Butchery, Paris, France, 1910, source: wikipedia
in a context dependent human behavior our responses can vary greatly depending on the situation. a human may be horrified by the boiled dog scene in one context (eg. a dinner party) but not in another (eg. in a survival situation).
ai and emotions
as ai continues to advance, researchers are exploring ways to make machines more human like, including the ability to experience and express emotions. this raises important questions about the ethics of creating ai that can feel and suffer, and how we can ensure that these systems align with human values and prioritize our well-being.
as of 2025, we now live in a world where LLMs can pass (in a very controlled environment) a turing test[1], also models can now use techniques like emotion recognition, generation and emotion-based decision making to look and sound more like a human[2], essentially llms can provide and interact with the usage of implemented primary human emotions. in contrary, emotions like empathy, envy, nostalgia, LLMs can mimic empathy, but not truly feel it, can express nostalgia, but not the bittersweet longing that comes with it.
image: primary human emotions first identified by psychologist Paul Ekman in the 1960s
let's imagine a conversation with an LLM where we're discussing a personal loss or struggle, here the LLM can generate a response that acknowledges your feelings and offers support but it's not truly empathizing with you. it might say something like: "i'm so sorry to hear that you're going through a tough time." while the LLM is using the right words, it's not truly understanding the depth of your emotions and the complexity of your experiences.
the dog banquet scene shows how humans can objectify and commodify other living beings, LLMs can also commodify human emotions in their responses. they can generate emotions like empathy, guilt, and nostalgia, but they don't truly experience them. instead, they treat emotions as something that can be created and manipulated at will in their responses.
one key difference between humans and LLMs is that humans have involuntary reactions to emotions, here the VK test subject's iris constricts when they are asked about their feelings, revealing their true emotions, it measures involuntary reactions (pupil dilation, blush response, heart rate) to questions designed to provoke empathy or distress, the idea is that androids, lacking genuine empathy, would not show these specific reactions. LLMs, on the other hand, can't exhibit involuntary reactions because they're not truly experiencing emotions rather can only mimic them with their pre-programmed responses.
the ethics of detection and categorization
if we use a VK-like tests to detect AI online, we need to think about what implications this has for how we treat and interact with AI entities. just as an assumption, would they be subject to special rules or even be treated as entities to be "discriminated", "eliminated" or "neutralized"? this also raises questions about how we categorize and understand non-human entities. by relying on a tool that measures emotional response, we may be reinforcing a binary distinction between human and machine, rather than exploring more nuanced understandings of intelligence and consciousness.
in the 1960's there was a research program in Canada called "fruit machine", the goal was the the research help the Government of Canada identify gay men working in the Public Service or to prevent gay people from obtaining government jobs.
this is a chilling example of how AI-like detection methods can be used to discriminate against marginalized groups.
this dark chapter in history should serve as a warning to us today. as we develop new AI detection methods, we need to be careful to not repeat the mistakes from the past and make sure that these methods are fair, equitable, well tested and respectful of human rights and we need to prioritize transparency, accountability, and human-centered design in AI development, so that we don't create new forms of bias and exclusion.
image: polygraph test, placement of detectors on subject. source: ebay
why police do nott just use lie detector tests to solve every crime, well, these machines don't actually detect lies rather only measure physical stress like a heart beat or sweaty palms. unfortunately, innocent people often get nervous and fail the test, while skilled liars can stay calm and pass it. because the results are so unreliable, courts in many countries refuse to accept polygraph tests as evidence. judges worry that a faulty test could send an innocent person to prison or let a guilty one go free. there is another point, the way the tests are read can be influenced by the testers biases. while they can sometimes help investigators narrow down leads, they are never considered proof of guilt. This is why the legal system prefers solid physical evidence over a machine guess.
in one of the dialogs, Rick Deckard, after the questionable VK result, mentions that he wants a bone marrow test of the subject. here this kind of test are legal in courts, the bone marrow test has been established as the definitive, scientifically irrefutable method for distinguishing humans from androids which implies that the VK test is merely a "psychological" guess, whereas the bone marrow test is a "biological" fact.
data collection conundrum
polygraphs collect deeply personal body data like heart rate and stress levels, now, to teach LLM to truly act and feel like a human, developers must feed it massive amounts of real human data, including recordings of our voices, facial expressions, and even our heart rates during emotional moments.
this creates a serious danger because the people whose data is used become targets for leaks or data broker data collection. data aggregators like palantir are already under scrutiny for security flaws, in 2024 and 2025, reports emerged that palantirs systems faced architectural risks where data could be accessed by foreign governments or leaked due to inherent design issues[3]. the company faced allegations of two separate data breaches in 2025 affecting consumer users, and ongoing controversies regarding its tools harvesting sensitive public records like medicaid data for government surveillance[4]. if polygraph data were fed into such systems, the combination of intimate biometric profiles with palantirs ability to merge vast datasets from health, finance, and law enforcement could enable unprecedented tracking, identity theft, or blackmail, turning a momentary stress test into a permanent, exploitable digital footprint.
tis is just a small example of how biometric date is being used without the consent. apart from palantir, many other companies like clearview ai, IDEMIA, NICE systems, cognitec provide this kind of data as a service.[5][6][7]
outro
i have finished exploring the Voight-Kampff test's ideas about AI and i think i reached a turning point. on one hand using this test to find artificial things sounds interesting. but on the other hand, it makes me worry about how we treat each other and machines.
in a world where machines can act like humans, we need to think carefully about what it means to be human. the VK test is like a warning sign that reminds us not to rely too much on technology.
this blog post was just a fictional idea I came up with, but it shows some scary truths about our relationship with AI today.