Collaborative study guide - Final (1).docx

Ex Directions: With your group, utilize the slides from that topic to note key concepts and key people (might be a scholar or a person representative of a case study) Also, note in a single sentence the main contribution of each of the readings/media from that topic Consult the midterm study g...

Ex Directions: With your group, utilize the slides from that topic to note key concepts and key people (might be a scholar or a person representative of a case study) Also, note in a single sentence the main contribution of each of the readings/media from that topic Consult the midterm study guide for suggestions in formatting Decision making (Group Gelman) Key Concepts Types of Reasoning: Deductive, inductive, abductive, practical Deductive reasoning reasoning such that truth of conclusion is guaranteed if premises are true Inductive reasoning - making probabilistic generalizations from observation Abductive reasoning - inferring the best explanation with the available data Practical reasoning - deciding the best course of action given beliefs and goals normative vs descriptive, normative is what you should do, descriptive is what you actually do Making decisions under risk vs uncertainty, risk: risk is present when future events occur with measurable probability (such as bets, lotteries) uncertainty: uncertainty is present when the likelihood of future event is indefinite or incalculable (this is much of life) Objective vs Subjective value - not exactly mathematical for humans - especially with money Expected Value Theory (EV) - normative decision making: involves calculating the expected value of various possible outcomes to determine the best course of action. More objective and monetary Expected Utility Theory (EUT) - normative decision making: involves calculating the expected utility of various possible outcomes to decide. More subjective and a better model for decision making under uncertainty Utility - a subjective measure of happiness or satisfaction from an outcome or action Homo economicus - economics theories describe a special human-like species, who individuals act as “perfectly rational” agents Limitations on rational behavior - planning fallacy - tendency to not consider the combined likelihood of unexpected events Prospect Theory - K Diminishing Sensitivity: implies that low probabilities are typically given more weight than they would receive using expected utility Key Readings Prospect Theory by Kahneman: In prospect theory one needs to know the state of wealth (like in utility theory), but also the reference state. Key aspects of prospect theory are the “adaptation level” (relative to natural reference point), diminishing sensitivity, and loss aversion. Thoughts on a Crisis by Ariely: Ariely analyzes the 2008 housing bubble and subsequent crisis by highlight the planning fallacy (people underestimate how long it will take to finish a task), the “butterfly effect” (events that happen to small groups of individuals have large effects down the road for everyone else), and “learned-helplessness” (the world gives us unpredictable punishments and we don’t have an explanation for what’s happening). How Do People Make Smart Decisions (Ted Talk by Gigerenzer): People need to use heuristics (shortcuts) and good intuition to make good decisions as people don’t actually calculate all possible alternatives. A Meta-Analysis of Decision Making (DM) & Attention in Adults with ADHD: People with ADHD have deficient/risky decision making as they are more loss averse than people without ADHD. Key People Tversky & Kahneman - sometimes referred to as the fathers of behavioral economics, as they demonstrate that the human brain relies on heuristics and biases in decision making. Essentially, they play a large role in understanding decision making by proposing theories like prospect theory and analyzing decisions under uncertainty vs risk. Gigerenzer - german psychologist emphasizes the use of heuristics and good intuition in decision making; mentions the success of Harry Markowitz who used the simple heuristic 1/n to allocate money into n number of funds Theory of Mind (Group Gopnik) Theory of mind: The capacity to understand other people by ascribing mental states to them. A theory of mind includes the knowledge that others' beliefs, desires, intentions, emotions, and thoughts may be different from one's own. Must be able to attribute false beliefs to have theory of mind. Remember Sally and Anne task: Anne puts block in box and leaves, then Sally moves the block to a basket. When Anne comes back in, where will Anne believe the block is? Somebody with ToM will recognize Anne’s false belief, but someone without ToM will point to the basket. Problem with Not Incorporating False Beliefs: A study was done in which a chimpanzee was shown footage of someone struggling with something and then was given the option to choose something from a variety of objects to help the person who was struggling. In the case of a person struggling with a door, the chimpanzee would choose the key; however, it was unclear whether the chimpanzee chose the key because it was aware of the persons inner desire for a key (ie Theory of Mind), or if it just associated the door with the key. Inverse Planning (Problem): You go backwards from seeing an action, to a plan, and to beliefs and desires. Generally, we decide to treat the object as a “rational agent”- their beliefs and desires lead rationally to a plan, which lead rationally to action. It is not that the beliefs, plans, or actions are rational. Universal bayesian problem: Combining Theory of Mind + Bayes + Utility Theory Dennett - Theory of mind is basically treating an agent as rational as a way to figure out the mental states the agent has Good math correlating to how we reason does not mean that math is how we reason. Multiple algorithms can add, and so can we. We don’t know which algorithm our minds go through. We can't determine which algorithms mirror humans’ algorithms. (essentially, just because we can get a computer to do something we do, we don’t know that is how we do it). Bayesian Food truck example: At MIT, there are food trucks that serve different types of food. We can infer someone’s preference for each type of food based on their actions when they are presented with certain trucks at certain places. Lower level of theory of Mind - Seems to develop before false beliefs Santos - Monkeys will approach a quiet box to take food when a researcher is not looking, but won’t really care as much between the two boxes if the researcher is looking Implies that the monkey is aware of the human looking at the box and therefore will have opinions about the food being taken Shari Liu Show infants a figure (circle) that spent more energy to access the triangle (that the circle wouldn’t expend for the square). When tested to see if they expected whether the triangle and the square was the figure’s preference, the infants expected the triangle Shown by looking longer at more surprising event (when the figure chose the square) Shows that ToM develops within the first year. Key Readings Dennett: true believers: Three ways of predicted the behavior of a system: Physical stance: Predictions based on physics. Design stance: Predict what a system will do based on its design Intentional stance: You assume something is rational, make an inference about its beliefs and desires, and then take a guess as to its most rational course of action to achieve its goals (you attribute beliefs to the system). Dennett argues that there is no difference, metaphysically, between you and a drier. A “true believer” is only an object that can be “reliably and voluminously” predicted by the intentional strategy. Jara-Ettinger: Naive Utility calculus: Humans assume that others make decisions to maximize utility, even from a young age. “​Human social cognition is structured around a basic understanding of ourselves and others as intuitive utility maximizers: from a young age, humans implicitly assume that agents choose goals and actions to maximize the rewards they expect to obtain relative to the costs they expect to incur. This ‘naïve utility calculus’ allows both children and adults observe the behavior of others and infer their beliefs and desires, their longer-term knowledge and preferences, and even their character: who is knowledgeable or competent, who is praiseworthy or blameworthy, who is friendly, indifferent, or an enemy. We review studies providing support for the naïve utility calculus, and we show how it captures much of the rich social reasoning humans engage in from infancy.” Emotion (Group Hopper) Happy sad angry disgust fear Two perspectives on emotion Experiencing - a matter of our own minds Perceiving - a matter of other minds Human cognition must be investigated at least at two levels (i.e., personal and sub-personal) and this can be done in particular by direct observation. Is emotion innate/hardwired or not? Innate: Universality Darwin’s study of facial expressions Paul Eckman: Agreement on certain expressions across cultures suggests a core set of emotions Similarity to animal models Experiments on emotions in monkeys Specialized brain regions Patient SP: damage in bilateral amygdala area associated with fear and avoidance responses → unable to feel fear Not innate Emotions can’t be localized to specific parts of the brain A lot of overlap One emotion can show brain activity in multiple areas All brain regions associated with emotion have some association with all emotions, with the level of association for each emotion differing by region Families, communities, and societies impart moral values, norms, and beliefs that influence our moral judgments and behavior. Universality in emotion by body gesture, not facial expression—tennis victory/defeat study Perceiving emotions Inferring emotions is an inverse problem—part of theory of mind Abductive Inference - best conclusion from incomplete data Another theory: We recognize emotion because we feel emotion ourselves. Explains why AI doesn’t understand emotion You need more information than just a face to determine emotion, like a body Why do we have emotions? Moral anger and deterrence anger and unsavory reactions deter people from “messing with you” Autism Spectrum Disorder patients express emotion in a way that differs Readings: Saxe + Houlihan: We attribute emotions to other people based on inferences about their goals, values, constraints, and actions. Essentially, perceiving emotions in others stems from abductive reasoning inferences (Bayesian belief updating) about what emotions should be associated with a given scenario. Frank (Theory of Moral Sentiments): Although emotions override rationality, their expression is evolutionary advantageous. Moral anger and emotion driven irrationality can help overcome problems of defection and commitment. Lisa Feldman Barret: We “construct our emotions” in that emotions are essentially predictions that our brain makes based on its experiences and other stimuli that it receives. Emotional Intelligence: the ability to manage both your own emotions and understand the emotions of people around you. Aviezer et al: Body cues, not facial expressions, are responsible for discriminating extreme positive and negative emotions—shown by tennis player’s victory/defeat pictures Belin et al.–children with autism may express emotion differently; experiment done with different shapes and what emotion(s) they associated with the shapes Cooperation (Group Kratzer) Humans cooperation differs from animal cooperation in forms of reciprocity, punishment, and fairness - and is a necessary feature of large scale societies in order for them to thrive Key Terms Reciprocity Fairness Punishment Direct reciprocity: Two individuals interact repeatedly: You cooperate today to earn your partner's cooperation in the future. (found in non humans) Indirect Reciprocity: involves reputation: My actions towards you also depend on your previous behavior towards others. (not found in non humans) Second-party punishment(found in non-humans) and third-party punishment(you don’t): punishment by a person who is directly affected, punishment by a person who is not directly involved Society scale and punishment: Smaller scale societies rely on direct second party punishment while larger ones use third party punishment. Disadvantage(found in non-humans) and advantage(not found in non-humans) inequality aversion - people react negatively and are averse to inequity to receiving less(disadvantage) or more(advantage) than others Thesis Cooperation is not just a quirky aspect of the human mind, but was necessary in forming complex societies Justification Dictator game Rand, et al. 2012 study (Reading) Shows that humans natural instinct is to cooperate The Farmer’s Fable and multiplicative growth Shows that pooling resources together will inevitably yield more products Moral anger and deterrence DICTATOR GAME (and its iterations) Regular Dictator Game: Dictator vs Receiver - humans are surprisingly generous Ultimatum Game: Ability to reject division of money - forms of punishment move cheaters away Iterated Stranger Version v Iterated Partner Version - People contribute at a higher level when playing with the same partner Third Party Punisher - An outside participant is able to spend their own money to punish the dictator - (⅔ of people punished infractions of the “norm”) Economic Games Across Cultures (Henrich et al. 2010) In smaller scale societies - there is often little third party punishment or any sort of indirect reciprocity In order for larger societies to work there needed to be enforced norms to hold others accountable despite a potential lack of connection Other important Ideas: Humans are not necessarily naturally cooperative but when given a short period of time they tend to be more generous than they have to be Humans can let emotion and punishment override rationality Third party punishment games are powerful tools for studying the characteristics and content of social norms. Morality (Group Rosch) Human morality arises from a complex interaction of reasoning, emotion, theory of mind, and probably some kind of innate structure (that we can deviate from with some initial learning) Key Concepts Prescriptive vs. Descriptive Morality 4 broad approaches to morality: Virtue ethics: Aristotle Person focused: do whatever a good person would do Critique: what defines a “good” person? Deontology: Kant Rule focused: do whatever is in line with moral rules + obligations Consequentialism: Bentham + Mill Outcome focused: do whatever maximizes utility Contractualism: key figure is Scanlon Group focused: do whatever we would (rationally) contractually agree to Trolley (switch) problem vs. Footbridge Why do they elicit different answers? Doctrine of Double Effect: It is permissible to cause a harm as a side effect of bringing about a good result even though it would not be permissible to cause such a harm as a means to bringing about the same good end More recent explanation: any difference is just people being irrational (do the math!) Basic idea: you are just being overwhelmed by your emotions in the footbridge case Learning right from wrong: Traditional theory: humans are taught Contemporary theory: innate structure in the brain that helps humans determine right from wrong Studies of morality in babies: when shown simulations of 2 shapes, one a “helper” + one a “hinderer,” when presented with the 2 shapes babies often reach for the “helper” Further study shows babies prefer “helpers” over neutral actors + neutral actors over “hinderers” Evidence for some sort of innate structure that guides morality → still an ongoing debate Accidental harm vs. attempted harm vs. intentional harm Crying baby case: Suffocate baby and you (all) are safe Let baby cry and you (all) die Is it morally acceptable to smother your baby in order to save yourself and the other villager? Kant/Deontology (no!) - certain moral rules and obligations we must follow Bentham/Mill (yes!) - consequentialism (maximize utility, for the greater good of the people) Innate ability: Key People Josh Greene: emotion leads humans to act irrationally; Dual Systems Theory of Moral Judgment Two distinct systems for moral judgment: emotional (automatic, unconscious) and rational (slow, conscious deliberation) Impersonal: working memory areas Personal: emotion areas Cannot actually understand “moral truth” based on human moral intuition because emotion influences our intuitive decision making Critique: the way the brain represents emotion → his study’s technique of mapping brain regions + looking for emotional activity is not necessarily a concrete way of representing the existence of emotion + their impact Phineas Gage: emotion is not enemy of moral judgment Koenigs: acquired psychopathy from PFC damage (evidence for innate structure) Readings Greene - fMRI Investigation on Emotional Engagement in Moral Judgement Found evidence through fMRI imaging for the hypothesis that personal dilemmas engage people's emotions more than impersonal dilemmas, and thus brain areas associated with emotion would be more active during contemplation of personal dilemmas Paul Bloom - The Moral Life of Babies Innate sense of morality (helpers and hinderers mentioned above) and desire to help others (results from spillover), but we are biased toward our own kind. Thus, a fully developed, unselfish morality relies on cultural development and rational insights. So how do we compare? (Group Treisman) Key Terms: Comparative cognition = study of cognition in other animal species Ethology = study of animal behavior; how adaptive behaviors have evolved over time Convergent evolution = species adapt similarly due to similar environmental pressures Number sense= ability to track expected number of objects Fast mapping= creation of a new label that is internalized forever after a single exposure Key Concepts: Biology only makes sense in the light of evolution Correlation between body weight and brain size, more neurons may result in more complex cognition ‘Core cognition’ of objects: humans and animals share many of the same core cognitive abilities (ex: numbers) Are spatiotemporally continuous Persist through property change Are bounded Nonhuman primate theory of mind: Can represent what others know and ignorance CANNOT represent belief → fail false belief tasks What makes humans special? Incorrect Theories: We have a better working memory Chimps do better on WM tests We can use tools So can crows (they can also change tools to better fit their purpose) We care about fairness Chimp got mad about cucumber vs grape Conscious vs subconscious processing: Monkeys can understand unconscious/subconscious cues We can learn from others Octopus learned how to open a box from observing other SYNTAX, BELIEF REPRESENTATION, and THEORY OF MIND are unique Key Animals: Chimpanzees “Proto theory of mind”: Acquire a sense of what people know, but cannot understand human beliefs (no false belief task) Have a sense of the spatiotemporally continuous property of objects, but unaware of the object persisting through property change Communication systems: vervet monkey calls, plausible, distinct sound or symbol that is intended to represent semantics (efficient way to alert a community is a case of convergent evolution) Nim Chimpsky: unable to acquire syntax, could not memorize hierarchical structure for sentences, no labels to distinguish actions, location, and events Dogs Can distinguish known words from unknown words in an acoustic signal, even when prosody is controlled (can distinguish between appraising and neutral prosody) Readings: Effects of social context and scrub jays: Jays were allowed to cache either in private or while a conspecific was watching, and then recovered their caches in private. Jays with prior experience of pilfering another bird’s caches subsequently re-cached food in new cache sites during recovery trials, but only when they had been observed caching. Jays without pilfering experience did not Results suggest that jays relate information about their previous experience as a pilferer to the possibility of future stealing by another bird, and modify their caching strategy accordingly Word learning in a Domestic Dog: Evidence for “Fast Mapping” Can associate words to specific objects For unfamiliar words, use process of elimination with a group of objects to associate the unfamiliar word with the unfamiliar object Doesn’t produce any type of syntax, but can understand a combo of verbs and objects Fast mapping is done through exclusion (new word associated with novel object) Suggests learning new words might be a general learning mechanism that can be found in other animals and not a language acquisition device that is specific to humans Rico (Border Collies) learned the name to more than 200 objects and was able to learn the name of novel objects through a process of “fast mapping” and exclusion with the same accuracy as a three year-old Mind of an octopus: Octopuses have evolved to be very smart Octopuses have an ability to adapt to the special circumstances of captivity and to their interactions with human keepers (they can recognize individual humans) Octopuses can quickly tell that something is not food but still want to play with it The majority of an octopus’ neurons are in their arms The arms are probably controlled by a mixture of localized control and top-down control Go! (AI I) and Artificial Intelligence (AI II) (Group Churchland) Key People: John McCarthy Co-authored the term “artificial intelligence” Marvin Minsky Computer scientist and cognitive scientist that was a pioneer in artificial intelligence Key Concepts: Return to three main principles of cognitive science materialism/naturalism Reduction computation Determinism: a complete specification of the state of the universe at time t, Incompatibilism vs. compatibilism If determinism is true, then we don’t have free will versus determinism does not rule out free will The cost of interpretability We dont know what AI understands because we dont work the same way The cost of compositionally/productivity Skydiving hot dog example The cost of generalization This is a result of what is known as “over-fitting” To what extent is the line following the data Dartmouth AI Conference The place in which the term “AI” was coined A gathering by top AI scientists regarding the future of the field The cost of using what you are given If an AI is built on “garbage Inputs,” it will “output garbage” An example of this was the Microsoft AI Twitter account @taytweets Worse are predatory models like “COMPAS” a felon recidivism model How AI works Neuron: Mathematical model representing a biological neuron; consists of inputs fed into a mathematical function (Neuron) that outputs a specific value; Function incorporates weights, biases, and activation function to determine output Hidden Layers: Refers to groups of neurons grouped together all fed output from a previous group of neurons; Backpropagation: Method of adjusting weights of each neuron to achieve more accurate results Teaching an AI to learn like a child: Readings: “The Dark Secret at the Heart of AI” There is no clear cut explanation of how AI systems make decisions, even for simple decisions such as recommendation engines on websites Deep Patient: An AI that was able to predict diseases and find hidden patterns in hospital data; an example of a powerful, unexplainable AI Deep learning is responsible for the explosion of AI AND their unexplainable nature DEEP DREAM: A Google experiment was used to determine how a Deep Learning NN recognized images using specific features; Feed the same image into a deep learning recognition algorithm and have the program modify the image Overtime, significant features used to recognize the image would be altered or would stand out Shows how deep learning is different from human perception and hone in on familiar visual features ”How researchers are teaching AI to learn like a child” Humans are born with ingrained instincts that help them think abstractly and flexibility “Causality, cost-benefit analysis, and types versus instances (dog versus my dog).” should be ingrained in AI system “We have at least four "core knowledge" systems giving us a head start on understanding objects, actions, numbers, and space. “ AI developers are programming systems that have zero ingrained rules, due to the bias “simpler is better” “Google's DeepMind has pushed deep learning to its apotheosis. Each time rules were subtracted, the software seemed to improve. Alpha Go Key people Lee Sedol - the world champion that was beat by AI. He had less than 0.25% of the training that the AI did. Key concepts Not able to generalize to other games. When alpha go was tested on a smaller board it was not able to beat a child A very fast game for a professional go player could take up to 9 hours Alpha zero is trained on 21,000,000 games Other versions: Muzero: A version of AlphaZero that can generalize to multiple games, including and especially Atari video games (Most recent version of original program) Readings The movie AlphaGo: talked about how they created the robot to beat the world champion Go player w.w Determinism: a complete specification of the state of the universe at time t, by the laws of nature, completely determines the state of the universe at all other times Incompatibilism vs. compatibilism Incompatibilism: free will is incompatible with determinism… Hard determinism: …so we don’t have free will Libertarianism: …so determinism is false Compatibilism: free will is compatible with determinism Paradox of freedom: We have free will We are not free if our actions are determined We are not free if our actions are undetermined Adina Roskies landscape of freedom The Libet experiment: is conscious will causally efficacious Aims to measure time of consciousness of willing an action relative to the time of neural signals related to action How do we define consciousness? Functionalism: mental states are defined by their causal roles, not material makeup Bottle opener example in class → each object is a bottle opener because of its function, not because of its particular design/appearance Explanatory gap → Brain + physical laws + ??? = consciousness Hard problem: How do physical processes in the brain give rise to subjective experience? Panpsychism: the idea that everything has a mind or mind-like quality Eliminativism: believing that we have a conscious experience when we actually don’t May need to give up on current scientific frameworks to solve Easy problem: Explaining how the physical systems work on a mechanistic level. Solved by materialist functionalism Readings: The Dark Secret -”How researchers are teaching AI to learn like a child” What does your brain have to do with it? (Group Kanwisher) Key Concepts The grandmother cell theory Specific neuron activates when grandmother is brought up, and then that message gets sent up to the brain and we think of our grandmother But it’s puzzling… Losing neurons doesn’t cause memory loss There are an infinite number of thoughts we can have but there are a limited number of neurons One potential perspective: It’s actually a neuron for a combination of features and these features are distributed across the brain It’s not actually a neuron for Grandmother → Grandmother is just a small part of a distributed representation So if you lose this neuron, others still represent the same info Human connectome project The goal of the human connectome project was to map the structures of our brain, but we can’t predict thoughts or actions simply based on neural structure. It is similar to seeing the hardwiring of a computer, just because you can see the system doesn’t mean we can predict the code that it is running. This project is helpful for understanding how the structure of our brain may change given age, cognitive diseases like dementia, or substance abuse disorders. C. elegans C. elegans is a tiny little worm with a very simple nervous system that is the exact same for all members of the species. We have mapped the entire nervous system for this animal (essentially, the goal of the human connectome project). Even though we’ve mapped their entire nervous system, we still can’t predict the behavior of the animal. This connects back Marr’s three levels of processing: computational, representational, and implementational. We might be able to understand the circuits and the behaviors, but we don’t understand the algorithm going on to cause certain behaviors/outputs. “Trying to understand perception by studying only neurons is like trying to understand bird flight by studying only feathers: It just cannot be done.” - David Marr fMRI fMRI isn’t the only way to measure cognitive processes but is the most popular, all methods must balance temporal resolution (time it takes to complete the scan) and log spatial resolution or essentially, how small and detailed a photo it can take. The fMRI is essentially the center point of all of these options and is thus the most popular. The dominant way mental processes are studied in the brain We don’t necessarily understand what our minds are doing by just looking at the brain But if we make the brain do something and measure what happens, we can (kind of) guess when it is doing that same thing again Like measuring the levels of pollution in L.A. and being able to (kind of) recreate the traffic patterns Can recover some information but not necessarily understand the kind of information being processed What it looks like: Brain “lights up” when engaged “Lights” that come on in the brain are actually a relative signal change What’s actually going on: Creates this magnetic field to get all the hydrogen protons aligned in a particular way Radio frequency pulse knocks of protons’ spins Brains return to normal spin, and speed of it recovering depends on tissue type Neural activity computed from recovery time Rely heavily on data processing and statistics to make sense of the “noise” Good because helps us overcome some of the inherent difficulties of measuring brain activity Bad because it can mislead us by allowing us to interpret noise BOLD (Blood Oxygen Level Dependent) response Local increase in cerebral blood volume during neuronal activity Use fMRI to measure the increase in the proportion of blood oxyhemoglobin Limitations The timescale of neural activity is on the order of milliseconds (firing rate) but the rate at which we measure in fMRIs is ~2 seconds Why don’t we measure more quickly? There’s an inherent tradeoff between temporal resolution and spatial resolution Would get worse spatial resolution if we measure more quickly Also, very sensitive to movement If you move your head, all of the blood in your head also moves More recent tool: multivariate analyses (MVPA) Approach we’ve been talking about (univariate analysis) tells us only which voxels were more or less active at one time vs. another Can see which information we can decode just from patterns of brain activity, so we don’t have to find an area that is sensitive to each specific thing Key people David Marr Return to the idea of the three levels of processing and ‘A Bridge Too Far’. Mostly relevant when thinking of the human connectome project and how mapping the brain won’t inherently help us predict thought or behavior. Hubel and Wiesel Experimenting on cats by showing different images to try to fire specific cells/neurons, they discovered when a really faint line was shown to the cat, neurons would activate (didn’t seem to activate for normal images or dots). Connected to the theory that center-surround cells in vision allow for edge/line detection - or the overall idea of neurons being specified for a task, or person (‘Grandmother Theory’). Nishimoto and Gallant These researchers recorded brain activity while the subject watched several hours of movie trailers, then they used regression models that translated between shapes in the movies and brain activity. They used these models to predict what movie trailers would look like using brain activity. Readings Weisberg Seductive Allure 3 Types of People (Extraneous Neuroscientific Info) 1: The Ignorant They said fairly consistently that the bad article was worse than the good article w/out neuroscience information They said fairly consistently that the bad article w/neuroscience info was worse than the good article w/neuroscience info They said that the bad article was dramatically improved w/neuroscience info and good article was improved w/neuroscience info 2: The Student Same as the ignorant 3: The Expert They said fairly consistently that the bad article was worse than the good article w/out neuroscience information They said fairly consistently that the bad article w/neuroscience info was worse than the good article w/neuroscience info They didn’t think that the bad article was improved w/neuroscience info and that the good article may have become even worse w/the neuroscience info Why the Contrast? “A plausible heuristic might state that explanations involving more technical language are better, perhaps because they look more “scientific.” “Placebic” information generally tends to increase compliance Pereria 2018 “Here we present a new approach for building a brain decoding system in which words and sentences are represented as vectors in a semantic space constructed from massive text corpora.” “In summary, we report a viable approach for building a universal decoder, capable of extracting a representation of mental content from linguistic materials. Given the progress in the development of distributed semantic representations, we believe that the semantic resolution of brain-based decoding of mental content will continue to improve rapidly. Our hope is that our work will serve as one of the cornerstones for developing and testing specific proposals about the nature of concepts, the organizing principles of the semantic space, and the computations that underlie concept composition.“ Nishimoto 2011 Took patients under an fMRI scanner and forced them to watch several movies. The scanner scanned brain activity (BOLD (Blood oxygen level-dependent) signals) in order to create fMRI images for a machine. The machine then used the patterns of brain activity and linked them with the movie scenes to use as data. The machine then was able to recreate (partially) the images in the movie using the brain information. Free Will and Consciousness (Group Carey) Key Concepts Basics of Cognitive Science: materialism, reduction, and computation These are how we make sense of the mind/brain relationship Determinism from a philosophical perspective Everything in the universe is determined, based on the full set of physical/natural laws and all of the “stuff” in the universe the Paradox of Freedom Incompatibilism: Free will and determinism are incompatible… … therefore, we don’t have free will (Hard determinism) … therefore, determinism is false (Libertarianism) Compatibilism: free will is compatible with determinism As long as decisions are made in the traditional decision making way, even though an individual’s beliefs and desires are outside of their control, the process of the decision making = free will (this is the standard stance of most philosophers) Illusion of Free will Priority: thought slightly precedes action Soon et al.: “the outcome of a decision can be encoded in brain activity of the prefrontal and parietal cortex up to 10s before it enters awareness” Libet (see below for more): there is an increase in brain activity before decision is made Classical interpretation of results: the brain has already decided your decision before your are consciously aware of it Alternative explanation of results: Stochastic Decision Models — the brain activity spike before the decision is the brain collecting evidence for the final decision Consistency: thought and action are conceptually related Exclusivity: thought is the only apparent cause of the action Functionalism: mental states are defined by their causal roles, not material makeup Analogy of bottle opener: can be instantiated in a number of different ways, but what makes it a bottle opener is the fact that it realizes the functional role of opening bottles Functionalism in consciousness: must be able to define what “something is like” Philosophical zombies: They look and act exactly like humans but in terms of their perception of the world there is “nothing it is like” to see a sunset or smell a rose, for example. They don’t have the human experience Examples (? how are they connected to the above?) Blindsight: visual information about the physical word is processed, but without consciousness (the person doesn’t realize they are processing visual information) (clip of visually impaired person maneuvering through hallway littered with mini obstacles) Anton’s Blindness: conscious experience of “sight”, but no actual visual information is processed (TV show clip of a hospital patient who insists they can see, but what they describe doesn’t match up with the physical world that the audience and other characters see) The “easy problem” of consciousness: detecting consciousness in the brain The “hard problem” of consciousness: we will never find a reductive explanation of consciousness in terms of physical properties. Subjectivity is a fundamental part of the universe. Some ways of addressing this problem: Panpsychism: everything is conscious, has subjective experiences, and can feel what something is like Eliminativism: consciousness doesn’t exist Eventually letting go of current frameworks: our current approach fundamentally cannot explain some facts about the world, such as why we experience things, then that's a reason to give up on our current approach. This results in an: Explanatory Gap: we know everything about the physical brain and the laws of physics, yet we cannot predict a person’s behavior → there is no reductive explanation available Key People Libet: Experiment where participant lifts finger while looking at a clock. Libet aims to measure time of consciousness of willing an action relative to time of neural signals related to action. Libet found that he could predict an action before the person had made a decision to commit that act. Thalia Wheatley and Daniel Wegner: Wrote about the illusion of free will– criteria of priority, consistency, and exclusivity David Chalmers: proponent of panpsychism, argues that consciousness is a fundamental property of matter (basically, rocks have conscious experience too) Dan Dennett: Proponent of eliminativism, argues in Consciousness Explained that we don’t actually have conscious experience Adina Roskies: the landscape of freedome Readings Unconscious determinants of free decisions in the human brain (Soon et al): The outcome of a decision is encoded in a person’s unconscious brain activity up to ten seconds before they consciously decide. Neuroscientific Challenges to Free Will and Responsibility (Roskies): Advances in neuroscience probably won’t undermine our societal views of free will and moral responsibility any time soon. Detecting Awareness in the Vegetative State (Owen): fMRI imaging may be used as a diagnostic tool for patients who are assumed to be vegetative, especially those who are unable to respond to traditional diagnostic cues. What is it Like to be a Bat? (Nagel): We currently have no way of representing the subjective experience of being conscious, which creates a problem for a materialist model of the mind. Directions: part 2 Can you think of examples to illustrate the following concepts? Decomposition: Theory of Mind multiple realizability Doctrine of Double Effect? Cost of interpretability Universal Turing machine Materialism/Naturalism: fMRI Paradox of Freedom? Libet’s EEG experiment Grandmother Cell Theory All the Animal Stuff Reduction: Grandmother Cell Theory fMRI Lesions to PFC alter moral behavior (change in mind results from a change in the brain) Transcranial Magnetic Stimulation Computation: Deep Learning AlphaGo DeepBlue Deep Dream We can’t predict behavior/understand consciousness from solely the physical breakdown of a system