Philosophy Essay: Against Representation

**This was for my ARTI/PHIL/PSYC 3550 class as my final essay and in summary, this essay argues against the classical models of cognitive science where representation is seen as essential to artificial intelligence. Final grade was a B+, so there are points where there might be thin spots, but overall, if y’all have any thoughts or points of disagreements, leave’em in the comments below. :)**

Introduction

Throughout much of our study and development of human cognition and its replication in the various forms of artificial intelligence, there has been an underlying assumption from which we have based our work on– that to replicate intelligent thought and intelligent behavior requires extensive representation. However, representation isn’t only unnecessary but that it would actually be detrimental to our efforts to create true artificial intelligence and our understanding of our own cognition if we keep the level of representation that we currently implement in our machines.

Embodied Cognition

“A machine can exist in an abstract plane where it can crunch numbers detached from the physical world. We, on the other hand, are enmeshed in it.”

A key difference between us and the machines we create is in our embodiment. A machine can exist in an abstract plane where it can crunch numbers detached from the physical world. We, on the other hand, are enmeshed in it. Our bodies constantly feed us information about the environment and we have no way of disconnecting from it. Our cognition is fundamentally embodied and our bodies affect the way we think and vice versa. 

This can be shown in our use of language. Language is a cognitive tool that both permits the expression of and the limitations of our cognition and how we use it tells us a lot about how we think. Since so much of our cognition is expressed and shaped within the constraints of our language, Language shows us just how much of our cognition is rooted in our bodies. For example, warmth is associated with affection (“warming up to someone”), weightiness is associated with value and importance, and exposure to immoral or unethical instances causes a feeling of uncleanliness [1]. In these cases, it shows that there is a reciprocal relationship between our cognition and our physiological form. 

This strong connection between body and mind means we often never have to completely hold our thought processes in our brains. For example, if we are working out a math problem, we can write it down and refer back to earlier steps to help complete later steps. This way, we only have to store the current step of the process in our brain, make the immediately relevant calculation and then the information can be stored and kept track of on an external medium. Since our brains are limited in its energy stores, it saves a lot of energy by processing information in small chunks and to externalise it like this. Rather than wasting time and energy replicating a model of the problem space internally to manipulate, we can just reference reality to inform us on what to do next. This ties back to cognitive technology [2] where external mediums can be used to bolster our cognitive processes and therefore, become an extension of our cognition. It makes sense, then, that since we can save energy, increase our cognitive abilities and make use of a body we were born with, a lower level of representation and a more embodied cognitive process will allow us to make better use of the body we possess and our energy resources and therefore, is a much more efficient way than the fully representational classical models. We will see how this need for efficiency means that, in our brains, just the minimal level of representation is used to allow us to “scrape by”.

Efficiency Argument

The human brain and its mysterious inner workings that we try to replicate in artificial intelligence have been shown to use far less representation than we thought. We can see this in all the ways that the brain fails to notice what should have been significant details like in the famous Gorilla Experiment [3] where participants, upon being asked to watch a clip of a basketball game, fail to spot when a person in a gorilla suit walks through the middle of the game. If the brain used the same level of representation as our machines, then all the participants should have made a complete internal model of the basketball game and noticed the gorilla. Instead, it seems that the brain is selective in its attention and by doing so, restricts the amount of information it needs to process at one time and is, therefore, more efficient.

” Rather, it seems that, for the most part, our brains can get by with as little representation as possible to be functional and be generally correct when it comes to problem-solving.”

Besides incomplete representation, the brain is also prone to a phenomenon called gist memory[4] where our memory can be, at times, approximative and at worst, unreliable. In a task where participants are asked to remember words with similar associations like “ice”, “snow”, and “winter” from a list, participants often say they remember a word, like “cold”, that wasn’t present on the list but also shared those associations. Other shortcomings like the notoriously unreliable eyewitness testimony have further exposed just how little our brain represents from the world. Rather, it seems that, for the most part, our brains can get by with as little representation as possible to be functional and be generally correct when it comes to problem-solving.

Evolution Argument

However, we are not the only ones capable of exhibiting intelligent behavior. In fact, much of intelligent behavior doesn’t need a brain but rather an interlocking system of simple operations that, when viewed gestalt, suggest intelligence. Roboticist Rodney Brooks coined the term “subsumption architecture”[5] to describe such a structure. We need to look no further than our own bodies for an example. Our immune system possesses remarkable adaptability and complexity[6] that, if we didn’t know any better, we might have thought that the individual cells within our body possess their own intelligence. From the outside, the operations of the cells that make up our immune system seem to be a conscious, coordinated effort but each cell is actually just responding to certain stimuli released by other cells like proteins or hormones. 

This sort of organisation can be seen in organisms like ants where they use very simple chemical markers that elicit simple behaviors that, when repeated within a colony of thousands or millions of members, can forage for food, wage war, maintain fungus farms and aphid herds, and build floating rafts to survive floods. Brooks himself built creatures based on crickets that showed how intelligent behavior like finding mates which require pathfinding and spatial reasoning can actually be the result of very simple physiochemical reactions within the body[5].

Therefore, not only is representation not needed at all for certain intelligent behaviors but even for the human brain, representation is only very minimally used. However, there is more to our intelligence than merely problem-solving and so far, it doesn’t seem like they’ve been accounted for. 

Rebuttals

Imagination, for one, hasn’t been accounted for. Imagination, according to Merriam-Webster, is “the act or power of forming a mental image of something not present to the senses or never before wholly perceived in reality”. This means that, necessarily, to be able to imagine means that the things we imagine must not already exist and therefore unable to be represented. It can be argued that our imagination is actually just a composite of previous experiences (representations) put together in novel ways to make something technically “new”. However, it must be considered that just because representations of something exists doesn’t mean that it’s necessarily real. For example, illusions and hallucinations causes a person to represent something that isn’t actually there. I am of the thought that the false representation of a non-actual object cannot be truly considered to be represented the same way we represent what’s real[see 7]. In the same vein, although I cannot yet explain the presence of these false representations, I argue that imagination, due to its focus on the non-actual, does not in fact rely on representations or at least, the type of representation as we currently understand it to be.

“We don’t know why it is so, it just feels right.”

Second, our memory, it might seem, also depends on representations. After all, memory is formed through the encoding of past experiences. However, I argue for limited representation, not the total elimination of representation. It is true that a certain type of memory (fact-based, non-phenomenal memory) may be represented but it still doesn’t account for other types. Muscle memory, for example, is formed through repetition of a movement and once formed, is accessed without conscious effort. This is different from the explicit content that makes up represented memories. Similarly, ability-based memory also seems to lack representation. For example, when teaching another to drive, it is hard to articulate the feeling of the car, how to know how far to turn or how far the hood extends past the steering wheel. Perhaps a better example can be found in our use of language. We can usually go about our day and communicate with no problem but once we are forced to slow down and explain the finer details of grammar or convention, we are often stumped. We don’t know why it is so, it just feels right. Efficiency once again plays a role here. By cutting out the thinking part of certain operations like driving a car or figuring out grammar before speaking, the brain doesn’t need to waste resources “reinventing the wheel” and rather, just knows that certain stimuli should entail certain reactions. You don’t need to think, “the light is green and green means go” before pressing the gas pedal. You do it automatically. This way, the process cuts out thinking and representing entirely and can go straight from sensing to reacting.

It was given for a long time that the representational theory of mind would be the basis on which we can produce higher cognition in our own creations but now we know better than to let that be the end-all-be-all. With a better understanding of the reciprocal relationship between our bodies and our cognition, studies that reveal just how little our own brains rely on complete representation, and the fact that not everything that looks intelligent is intelligent, it seems that representation is actually a rather inefficient and insufficient explanation for all the miracles we are capable of.

References

[1] McNerney, Samuel. “A Brief Guide to Embodied Cognition: Why You Are Not Your Brain.” Scientific American Blog Network, Scientific American, 4 Nov. 2011, blogs.scientificamerican.com/guest-blog/a-brief-guide-to-embodied-cognition-why-you-are-not-your-brain/.

[2] “Chapter 9: Extended Minds?” Mindware: an Introduction to the Philosophy of Cognitive Science, by Andy Clark, Oxford University Press, 2014, pp. 192–211.

[3] Simons, Daniel, director. Selective Attention Test. YouTube, YouTube, 10 Mar. 2010, www.youtube.com/watch?v=vJG698U2Mvo.

[4 ]Makin, Simon. “What Happens in the Brain When We Misremember.” Scientific American, Scientific American, 9 Sept. 2016, www.scientificamerican.com/article/what-happens-in-the-brain-when-we-misremember/

[5] Brooks, Rodney A. “Intelligence Without Reason.” The Artificial Life Route to Artificial Intelligence, 2018, pp. 25–81., doi:10.4324/9781351001885-2. 

[6]Chaplin, David D. “Overview of the Immune Response.” The Journal of Allergy and Clinical Immunology, U.S. National Library of Medicine, Feb. 2010, www.ncbi.nlm.nih.gov/pmc/articles/PMC2923430/

[7] Loar, Brian. “Transparent Experience and the Availability of Qualia.” Consciousness and Meaning, 2017, pp. 273–290., doi:10.1093/acprof:oso/9780199673353.003.0016.

Full Experimental Design for Testing Cognitive Ability Through Temperature Control

*This Experimental Design was a project for an AP Psych class so it’s not representative of what a real experimental design looks like. This design also ties into a previous article on the site that goes over tbe subject of the study a little less formally. This design was based off of said article, actually. Click here to see the original article.

Section I: Proposal

Abstract:

There is a widespread myth that humans use only ten percent of their brains and how it implies that we have a lot of untapped potential. While the myth has been scientifically proven to be false, the fact that the brain usually isn’t usually functioning in its highest capacity might be true. There are concerns regarding the impacts of external factors on the brain’s performance in certain activities, like temperature. In order to test if there are effects on the brain in different temperatures and what the potential outcomes are, we are researching this topic and conducting this experiment to see if adult individuals placed into different temperatures will yield significantly different test performances.

 

Section II: Background Research

The human brain accounts for 25% of the body’s total glucose utilization and 20% of its oxygen consumption. While using these biological fuels for its cognitive processes, the brain also generates a lot of heat. However, to the fragile neurons in the brain, any significant deviance of the temperature from the baseline would cause incomparable damage. We also know that the brain has a maximum temperature limit, but its minimum temperature limit is not yet defined, meaning that the brain ails under high heat while it is more resistant under cooler temperatures. These points are supported in the naturalistic observational study conducted by Kiyatkin, E. A. (2010), where it stated, “Brain cells are exceptionally sensitive to heat, with some irreversible damage starting to occur at ~40°C, only about 3°C above normal baseline, and progressing exponentially with slight increases above these levels.”

While our brains are liable to overhead, we have also evolved ways to maintain the temperature. In “The Rationale for Human Selective Brain Cooling”, the authors B.A. Harris and P.J.D. Andrews (1970, p. 738) listed three ways in which humans have adapted to help maintain our brain’s temperature, “[the] cooling of venous blood by the skin which in turn cools the arterial (carotid) blood supply to the brain; cooling by heat loss through the skull; and cooling by heat loss from the upper airways.” With so many adaptations designed to cool the brain, then it is evident that keeping the brain’s temperature as close to the baseline temperature as possible is crucial to our survival. It also means that because of the amount of heat being exchanged at the level of the skin from the brain, then the room temperature of a location will affect the brain temperature of everyone present. Harris, B.A. et tal.’s paper was based on anatomical study and it also goes on to say that, “as the brains of early humans grew in size, their emissary veins developed in tandem, which supports the radiator theory that as humans evolved and developed bigger brains they must have developed an increased venous cooling capacity.” What that means is that by using existing data about our evolutionary history, it has been concluded that as our brains evolved to become more complex and more adept at higher-level thinking, as evolving bigger brains means the development of the limbic and the mammalian brain and therefore the development of higher-level thinking, our species developed more sophisticated cooling systems in tandem. In other words, as our brains began developing higher-level cognitive abilities and our brains began using up more energy and generating more heat from those abilities, we needed more effective and complicated cooling systems to keep up the level of thinking we were doing, therefore demonstrating the importance of the temperature of our brains in our level of cognitive ability.

There was an experiment done by Haider, B. et tal. (2010) that concluded that there were inhibitory neurons in our brain that stopped the brain from consciously processing information that is deemed irrelevant in an effort to save energy. In other words, because the brain is trying to conserve energy and limit heat output, it is also cutting back on what sensory information it is processing and committing to memory. This means that most of what we see and hear etc through our five senses and most of our more fleeting thoughts are lost before we are even aware of them. If the brain were to be in a cooler environment, then perhaps it won’t need to cut back on its processing power as much and our cognitive ability will grow as more information will be consciously available to us.

These studies all support the fact that the brain does, in fact, benefit from lower temperatures to offset the amount of heat it generates. While these studies don’t strictly show that cognitive ability is linked to temperature, it provides a strong foundation in which we will base our study on in proving the relationship between a cooler room temperature and higher cognitive ability.

Our study will be an experiment where we will administer creativity tests that test the participants’ cognitive creativity in the form of spontaneous drawings with nudges in the form of abstract shapes or lines. We will also administer cognitive tests similar in nature to CogAT to test cognitive ability in pattern recognition and spatial awareness as well as other high-level mental processes. The participants would be tested first in a room-temperature room to establish their baseline score before being tested again in their experimental rooms with varying temperatures. Our conclusions will be drawn from the difference in the participants’ scores from each individual’s first round of testing to the second, therefore their improvement or decline in test performance would be calculated in relation to each individual’s original score. By doing this, we hope to see if the temperature of a room affects cognitive ability.

 

Section III: Experimental Design

Our hypothesis is that if an adult is placed in a room with a cool temperature at sixty-three degrees Fahrenheit, they will have a higher level of cognitive ability and perform better on cognitive tests and creativity tests compared to their control data in a room-temperature room at seventy-three degrees Fahrenheit. However, if an adult is placed in a room with a hotter temperature at eighty-three Fahrenheit, they will have a lower level of cognitive ability than when they were in the room-temperature room and will perform more poorly compared to their control data as the participants who were tested the second time around in the cool room. The colder room, in turn, would lead to a boost in each of the experimental group’s scores on the test more than the hot room in comparison to their score when the participants were all in a room-temperature room.

In our experiment, the independent variable is the temperature of the rooms that each participant goes into to take the test (either the sixty-three degree cool, the seventy-three degree room temperature or the eighty-three degree warm room), while the dependent variable is the test results of the randomly selected adults. The independent variable would be the room temperature and the dependent variable will be the difference between the test scores of each participant’s control score and their experimental score. (improvement, no change, or decrease in scores).

To make sure that the data collected from the experiment was really dependent on room temperature and was not up to chance, we have to have a control group that will test both times in a room-temperature room (seventy-three degrees Fahrenheit).

To control for confounding variables, around several day’s time will be given to adjust for potential jet lag after the randomly selected adults travel from various locations in America to the place where the experiment will be conducted. The participant’s overall level of health, with factors like sleep, diet or stress level, will be monitored and kept with little to no variation between the two testing periods. Besides the room temperature, every other aspect about the rooms will be the same for every room so that distracting posters, noise, or other such things will not be present to affect concentration when the participants switch rooms. The participants are also all going to wear the same clothes in both rounds of testing to ensure that their body temperature is consistently being affected by the temperature of the room. The participants will take their first test and their second test at the same time of day so that the impact of different cognitive levels at different times of day for each participant will be nullified. The participants will also be provided with adequate water for both rounds of testing so that another important factor of brain functionality, hydration, is being controlled for. Different tests will be administered in each round of testing so that participants do not get the chance to be more familiar with the test in the second round. The difficulty of the two tests being administered for both rounds of testing will be maintained to ensure that the scores aren’t affected by an easier or harder test in the second round. Of course, we will make sure that participants aren’t affected by other participants by testing each of them separately. There is a time limit of one hour for the cognitive tests and the tests will be made so that it can reasonably be completed in this time. This is so that everyone has the same time limit and it will be made clear to participants that they are under such a limitation. The time limit for the creativity tests would be thirty minutes so that their creative ideas have to have some amount of real spontaneous creativity rather than let the participants ruminate on what they could draw. The time limit would also curtail the body’s ability to become acclimated to the rooms’ temperatures and would therefore render the difference in room temperature null. To ensure that all participants start off at about the same body temperature, all participants would be required to stay in a room-temperature room for thirty minutes before testing. There will also be small reward for participating in the study to ensure participants’ continuing levels of motivation.

This will be a single blind study, in which the participants will not know why they are being tested or what the variables the scientists are testing are, they will only be aware that they will take the tests right before testing begins. We will also control for experimenter bias by ensuring that multiple scientists facilitate the experiment in order to be sure that no one scientist manipulates the experiment to procure the hypothesized results or interprets the results wrongly simply to prove the hypothesis correct. We will conduct the experiment with participants selected randomly from around the United States in order for the results to be representative of American adults. The random sample will include participants who live in different states and they are going to be gathered together for the test. There will be thirty participants in total with fifteen being male and fifteen being female. For the two experimental groups (hot and cold) and the control group, the participants will be randomly selected to be in any of the three groups. In the experimental round, there will be five males and five females who will test under each of the room temperature conditions to ensure fair representation of genders, all randomly selected, of course.

 

Section IV: Ethical Concerns and Practical Applications

There are five key components for an experiment to be considered ethical. First informed consent is need from every participant of the experiment, meaning that they are given enough information about the experiment to understand the risks but not enough that it would significantly affect the outcome of the study. If the possible participant is under eighteen years old, then parent permission is needed. Voluntary withdrawal is another key component; the participant should have the right to stop participating in the experiment if they desire to. The third thing that those conducting an experiment should do is protect all participants from emotional or physical harm. Fourth is confidentiality, identities must not be revealed. The last component is debriefing; the experiment should be fully explained to the participant after it is over.

Any scientist or other individual/group must follow this guideline in order to conduct an ethical experiment. The experiment that we would conduct does conform to all the APA ethical guidelines. The thirty participants in the study would all be informed of the objective of the study and decide to give consent. It would be made known before the experiment started that they can withdraw at any moment. Neither will any of the participants be physically or emotionally harmed; they would be given the task to complete a creative and a cognitive ability test in different temperature rooms. The identities of the thirty participants would not be revealed at any given time, and they would be fully debriefed when the experiment is complete. 

If completed, this study would benefit the human race. This experiment could provide insight into ways that employers and educator could use to enhance cognitive ability, improving the education that children in schools receive and the productivity of the workplace. If it is proved that temperature affects cognitive ability or at what temperature the brain functions at the best then it would enrich the human race. It could lead to further knowledge of the brain and change the temperatures in school buildings, work environments, public establishments, etc. so that the individuals spending time in these locations work and perform at their maximum potential.

 

Bibliography

Harris, B. A., & Andrews, P. J. (1970, January 01). The Rationale for Human Selective

Brain Cooling. Retrieved August 29, 2017, from https://link.springer.com/chapter/10.1007/978-3-642-56011-8_66

 

 Kiyatkin, E. A. (2010, January 01). BRAIN TEMPERATURE

HOMEOSTASISPHYSIOLOGICAL FLUCTUATIONS AND PATHOLOGICAL SHIFTS. Retrieved August ccc29, 2017, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3149793

 

 Haider, B., Krause, M. R., Duque, A., Yu, Y., Touryan, J., Mazer, J. A., & Mccormick,

D. (2010). Synaptic and Network Mechanisms of Sparse and Reliable Visual Cortical Activity during Nonclassical Receptive Field Stimulation. Neuron,65(1), 107 121. doi:10.1016/j.neuron.2009.12.005