Agent Plus Environment

A new semester…

Long time, no writing — it’s the start of a new semester (my last semester!) and I’ve been busy with a number of different things:

1. The VC Women’s Fencing team. We’re in full competition mode. We conquered in Cleveland recently, vanquished difficult foes at Brown University, and are gearing up for a big match at Wellesley next Sunday, which will decide whether we claim the Northeast Conference Championship this year!
2. My undergraduate cognitive science thesis. I’m looking at the emergent behavior of a group of simulated prey robots that can communicate with each other about the presence of a predator. I have questions about communication, environment, and motivation. Being a year-long project, I’m supposedly halfway through, though in reality, it’s not so clear-cut. I spent all summer reading papers and doing background research, filled last semester with hypotheses, possible architectures, and more background research, wrote up a first draft this winter break, and am now hard at work on the simulation itself.
3. Taking photos of the weather. An unusually large amount of snow has fallen at Vassar — what better to do than document it with a camera? (Click for larger versions.)


Sunset Lake II



Snow Forest



Winter Picnic

4. Figuring out my post-graduation life. On the advice of many folks, I’m not heading immediately to grad school. My enthusiasm for learning, research, and knowledge hasn’t vanished — quite the contrary. I’m going to spend at least a year exploring the places outside the classroom, longer depending on where I end up. Academia-land? The wide world beyond? Still up in the air.

Rock on

Are you familiar with perceptual control theory? If you aren’t, the basic idea is this: People are not rocks. As Philip Runkel puts it,

“Living creatures behave very differently from lifeless things. Unlike a rock, a human does not just sit until something bumps it.”
– Philip Runkel, “Casting Nets and Testing Specimens,” pg 75

The idea is, organisms and agents and people get a bunch of different sensory inputs. They have some internal standards for what they want that set of sensory inputs to be like — some desired state of the world. The difference between how they want the world to be and what the world is actually like drives what they do — what we see as behavior.

The reason this is appealing to me? Perceptual control theory (PCT) says we’re not just input-output machines. Behavior is goal-directed and purposeful.

It’s a useful theory if you want to figure out why people are doing what they do and how to avoid or mediate conflict. Everyone has internal standards that they’re trying to control. As Runkel says,

“[M]ost of us very often act as if we expect other people to behave like rocks. And when we act toward other people as if they were rocks or blankets or typewriters or teacups, we make unending trouble for ourselves. It is true that people do have some features in common with rocks and typewriters. There are, however, important differences between living and nonliving things that most of us overlook time and time again, and to our sorrow.”
– Philip Runkel, “People as Living Things; The Psychology of Perceptual Control,” pg 14

If you want to learn more, I’ve found you a nice list of articles, an informative Less Wrong post a friend linked me to, a comprehensive website, and Google.

And yes, talking about PCT really just was my excuse to share those lovely quotes from Runkel.

Emotional intensity and the individual

Let’s say you’re at home. Maybe you’re lounging indolently on the couch, feet up on the brown wood coffee table, television whining at you from across the room. Maybe you’re cooking tonight’s dinner, chopping vegetables with careful strokes, sliding the ever growing pile of peppers and onions and tomatoes into a hissing frying pan. Maybe not. Maybe you’re in another room when the fire alarm sounds, bleep bleep bleep, blaring its cacophonous melody into your generally peaceful home.

How do you react?

Do you scream? Do you calmly turn off the stove, flap a towel at the cloudy air around the smoke detector, and wait patiently for it to detect that there’s not actually a fire? Do you leap up from the couch, tripping over the coffee table in your panic, terrified of burning to death in your own living room?

The strength of your emotional response to this (or any) emotional stimulus is known as emotional intensity. Emotional intensity can be measured with psychological scales, such as the aptly-named Emotional Intensity Scale (EIS) developed by Bachorowski & Braaten (1994) [PDF]. The underlying if obvious assumptions of these scales are that some individuals experience all of their emotions more intensely than other individuals, and all individuals may respond with different strengths to the same stimuli.

Your personality influences your experience of emotions

You may already be familiar with the Big 5 personality factors: Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism (sometimes called Emotional Stability). (If not, look them up.) Robert McFatter, in his 1998 paper Emotional Intensity: Some components and their relations to extraversion and neuroticism [PDF], investigated the relation between temperament and the intensity of positive and negative emotions. (Positive emotions included happiness and pleasure; negative emotions included worry, guilt, anger, and sadness.) McFatter described and tested several models, all of which had slightly different predictions about how neuroticism, extraversion, and positive and negative emotional intensity are correlated.

1. Larsen & Ketelaar model: The measures used to examine emotional intensity in this model tapped frequency of experienced emotions more than the intensity of single (and possibly infrequent) reactions. The model predicts that Extraversion is positively related to positive intensity and unrelated to negative intensity, and that Neuroticism is unrelated to positive intensity and positively related to negative intensity.
2. Larsen & Diener model: This model draws on the theory that the intensity of experienced emotions is used to regulate arousal levels. Arousal level can be tied to Extraversion, so this model predicts that Extraversion is positively related to both positive and negative intensity. Larsen & Diener also predict that Neuroticism is similarly positively correlated with positive and negative intensity.
3. Wallace, Bachorowski, & Newman (WBN) model: Extraversion is suggested to reflect a behavioral approach system and a behavioral inhibition system. Neuroticism is suggested to reflect the reactivity of an arousal system responding to the behavioral approach/inhibition systems that serves to prepare the individual to respond. This model accordingly predicts that Extraversion is positively related to positive intensity and negative related to negative intensity (and thus that Extraversion is overall uncorrelated with overall emotional intensity), and that Neuroticism is positively related to both positive and negative intensity.
4. Gray’s model: This model predicts that the behavioral approach/inhibition systems form dimensions that are rotated roughly thirty degrees from the Extroversion and Neuroticism dimensions, so they don’t line up. The model predicts that Extraversion is positively related to positive intensity but only weakly negatively related to negative intensity. Similarly, Neuroticism is predicted to be weakly positively related to positive intensity, and positively related to negative intensity. Gray’s model, furthermore, suggests that the negative emotions can be subdivided into anger/panic and anxiety/fear categories. These subcategories may have different relations to Extraversion.

Methods, Correlations, Analyses, Results

To test these models, McFatter gave a series of questionnaires to 1553 college students taking introductory psychology classes (596 male). Participants completed the 30-item EIS to examine positive and negative emotional intensity (14 items and 16 items, respectively), the Eysenck Personality Inventory (EPI) for measuring Extraversion and Neuroticism (in addition to subscales for impulsivity and sociability), and a third unrelated questionnaire.

Based on an initial factor analysis of the EIS, negative intensity was separated into two groups: anger/frustration (hereafter referred to as “anger intensity”) and non-anger, such as worry, guilt, and sadness (referred to as “non-anger intensity”). This result supports Gray’s theory that two separate negative emotion systems exist.

Consistent with both Gray’s model and the WBN model, Extraversion was shown to be positively related to positive emotional intensity (r=0.19, P<0.0001), negatively related to non-anger emotional intensity (r=0.18, p<0.0001), and unrelated to anger intensity (r=0.02). In plainer terms, individuals with high Extroversion scores tended to experience more intense positive emotions and less intense negative emotions.

Neuroticism, on the other hand, was shown to be positively related to all three kinds of emotional intensity, though less strongly to positive intensity (r=0.18, p<0.0001) than to non-anger or anger intensity (r=0.56,p<0.0001 and r=0.45,p<0.0001, respectively). That is to say, individuals with high Neuroticism scores tended to report experiencing more intense emotions overall. This is consistent with Gray's model.

A couple other interesting results:

Females reported significantly higher emotional intensity than males overall, with the largest difference seen in negative intensity (0.411, p<0.0001).

The positive relation between Extraversion and emotional intensity was stronger among people with a high Neuroticism score.

Neuroticism and emotional intensity

It’s hard to tell without reading a pile of psychology papers, but the fact that Neuroticism was positively related to positive emotional intensity was surprising. Previous results found a negative relation, though several of these had measured emotional intensity with a different scale–one that seemed to confound frequency and intensity of the experienced emotions. The WBN model, relatedly, claimed that Neuroticism reflected general emotional reactivity. (Recall the personality factor’s other name: Emotional Stability.) So McFatter investigated.

He found that when looking at the difference of the positive intensity and negative intensity scores, the relative emotional intensity was negatively related to Neuroticism, as in those previous studies. However, when examined on their own with the other variables controlled, the relations of both positive and negative intensity to Neuroticism were positive. The WBN model only explained a portion of the story.

McFatter’s results, overall, support Gray’s model and the WBN model, suggesting that the variations in positive and negative emotional intensity may be the result of separate emotion systems, but that they do have some common variation that may best be explained by their relations to Neuroticism.

References:
McFatter, R. (1998). Emotional Intensity: Some components and their relations to extraversion and neuroticism. Person. individ. Diff., 24(6): 747-758. [PDF]

Unresolved

I’m not one to make New Year’s resolutions.

I mean, sure, I could take my pick of popular New Year’s resolutions; I could decide, on the first day of the new year, that this year, I’ll start exercising more and eating better, or that I’ll spend more time with my family and friends, or that I’ll learn a new skill. And if I chose to make resolutions, I’d be far from alone–a 2008 survey on Dorthy.com found that 66% of the 2000+ adults polled had made resolutions at some point (though only 17% managed to keep them).

Making resolutions: It’s about self-control

The question Anirban Mukhopadhyay of the Hong Kong University and Gita Venkatarmani Johar of the Graduate School of Business at Columbia University asked is this: What determines how many goals a person will set, and how successful a person will be at achieving those goals? They performed a few studies in 2005 to look at the relationship between self-control, goal setting, and goal achievement. They suggested that what you believe about self-control affects the goals you set and achieve [PDF].

In the paper, “self-control” is used to mean a sense of willpower. Mukhopadhyay & Venkatarmani discuss various lay theories of self-control, noting that the amount of self-control a person has can be seen as either an inherently limited or unlimited resource, and that this resource can be seen either as malleable or as fixed (the amount of self-control a person has can change over time, or not). An important premise to note here is the idea that the probability of choosing a goal or making a resolution increases if a person thinks that goal can be attained. So if you think you’ll be able to achieve a goal, you’re more likely to set it. Combine this with theories of self-control, and in general, if you believe you have unlimited stores of self-control, you’ll set a larger number of goals. If you believe self-control is malleable but limited, you’ll set fewer goals.

Mukhopadhyay & Venkatarmani also discuss self-efficacy: belief in one’s capabilities, the perceived ability to carry out a desired action. They propose that people with high self-efficacy–people who believe that failure is the result of insufficient effort, and thus exhibit increased commitment and persistence–will achieve more of their goals than people with low self-efficacy, who tend to view failure as the result of deficient ability, and thus may simply give up.

The studies

In the first study, 85 participants (all college students) each read one of four passages presenting lay theories of self-control. Each passage contained two paragraphs; the first discussed self-control either as limited or as unlimited, and the second discussed self-control as either malleable or fixed. The participants then answered questions about their belief in each of two theories presented, followed by a second questionnaire to assess motivation, in which they listed all their current goals.

The study was testing whether a belief in unlimited, malleable self-control would result in most resolutions, and indeed, this is what was found. The experimenters had some concerns about participants’ natural beliefs in relation to the passages they read, however, so in study two, the order of the two measures (lay theories and motivation/goal listing) was varied. Data from 130 new participants revealed that, as hypothesized, if the motivation & goals questionnaire were assessed first, then among the people who believed self-control is malleable, those who also believed self-control to be unlimited (vs. limited) set more goals. When lay theories were assessed first, this result reversed. The people who believed that self-control is fixed were unaffected by order.

The third study moved on to examine goal achievement, adding a measure to look at self-efficacy. The study had two sessions, in November then February. In the first session , the 159 participants read passages about lay theories (much like in study one, but with longer passages to strengthen the manipulation), listed the resolutions they were planning on making at New Years, rated how disappointed they would be if they failed to keep their resolutions, and filled out individual difference measures (which included a self-efficacy scale). Only 86 participants successfully returned for the second session, during which they indicated how much success they had had at keeping their resolutions.

What does this mean for your resolutions?

The resolutions made by participants across all conditions were qualitatively similar (take a look at any list of popular New Year’s resolutions, and you’ll see the majority of the goals). As shown in the first two studies, more goals were set by people who believe self-control is unlimited and malleable than by any other people–that is, if you expect more success, you may increase the difficulty and number of tasks that you set for yourself. Self-efficacy did not have a significant effect on goal-setting.

As far as success goes, only the interaction between lay theory and self-efficacy was significant. If participants believed in limited self-control and were low in self-efficacy, they tended to give up more often, failing to achieve their goals. But if participants believed in unlimited self-control, self-efficacy had no effect; participants achieved just as many goals regardless, and people who set more resolutions were marginally more likely to succeed.

Mukhopadhyay & Venkatarmani realize that their research does not directly look at the relationship between lay theories of self-control and beliefs about one’s own amount of self-control and self-efficacy, and propose this as an area for future study. But in general, lay theories about self-control can determine how much success you’ll expect (and thus, how many goals you’ll set), and self-efficacy beliefs can determine how much success you’ll actually have.

References
Mukhopadhyay, A. & Johar, G.V. (2005). Where There Is a Will, Is There a Way? Effects of Lay Theories of Self-Control on Setting and Keeping Resolutions. Journal of Consumer Research, 31, 779-786 [PDF]

Billions and trillions

Step by slow, supercomputed step, we approach singularity.

This step: Two massively parallel cortical simulations, run at the Lawrence Livermore National Labs by Rajagopal Ananthanarayanan, Steven Esser, and Dharmendra Modha of the IBM Almaden Research Center, and Horst Simon of the aforementioned labs–these are the guys who previously simulated at the scale of mouse and rat cortices. They used a Blue Gene supercomputer (with a whopping 456 CPUs and 144 TB of main memory–just wait, ten years from now I’ll look back on this sentence and laugh at how little computing power and memory that is). The first, and larger, simulation included 1.6 billion neurons and 8.87 trillion synapses. Human brains still dwarf these numbers: roughly 20 billion neurons and 200 trillion synapses. But it’s a cat-sized step with the complexity and scale of a feline brain.

The first simulation used experimentally-measured gray matter thalamocortical connectivity from a cat’s visual cortex–the simulations neurons were connected in a biologically plausible fashion. Phenomenological spiking neurons, individual learning synapses, axonal delays, and dynamic synaptic channels were all included in the software. The second simulation, with 900 million neurons and 9 trillion synapses, had probabilistic connectivity.

Speed-wise, the researchers report that their simulation runs 2-3 orders of magnitude slower than real-time, when compared to a human cortex. With near perfect weak scaling (doubling the memory resource doubles the model size that can be simulated), human-scale models may be just around the corner… well, relatively speaking; the researchers predict it’ll happen in less than ten years. Just as soon as there’s a supercomputer super enough.

The research paper is also available at researcher Dharmendra Modha’s blog [PDF].

But bigger isn’t necessarily better

We may have to wait ten years for human-scale simulations, but we may not need a human-scale platform to be able to build intelligent AI. Researchers at Queen Mary, University of London suggest that bigger may not necessarily be better, when it comes to brains. A lot of complexity can be found even in tiny insect brains. Maybe it’ll be a swarm of honeybee robots that takes over the world!

The complexity of models

For a time, I was convinced that every model out there would not be an adequate model of what a human brain could do because every model out there had to simplify, and thus, that no model or computer software would ever be able truly intelligent until we had the computing power to make an electronic human. I knew there was value to models, but deep down, I retained the conviction that no model, no simulation, no AI would ever manage the same level of complexity or intelligence as a human without being, simply put, a human.

Fortunately, I was relieved of this notion around the same time I started taking Cognitive Science classes: Humans aren’t the only intelligent creatures, the point of a model is not to create the thing you are modeling, all models simplify some aspect (it’s just a matter of choosing which aspects are most important to get exactly right). The world may be its own best representation, as Rodney Brooks so aptly said, but that should not preclude us from simplifying the world to better understand how it works, nor should that, in return, prevent us from trying to simulate ourselves in software.

I, for one, am looking forward to watching the intelligent honeybee robots and the supercomputer human brains band together to overthrow the government.

Beautiful and good to eat

Deep down, maybe we all know we are, every one of us, a unique snowflake. But a lot of people, they don’t want it to be true. They want all the snowflakes to melt together into one big puddle. They want to be able to share their subjective view of the world with everyone else. They want to be able to look at a sunset and know that what it’s like for me to see the sunset is the same as what it’s like for you to see the sunset.

Hey, we all want things we can’t have. And in this case, science says no! Here’s a piece of wisdom from David Brin’s sci-fi novel Kiln People:

“We may use similar terms to describe a sunset. Our subjective worlds often correspond, correlate, and map onto each other. That makes cooperation and relationships possible, even complex civilization. Yet a person’s actual sensations and feelings remain forever unique. Because a brain isn’t a computer and neurons aren’t transistors.

It’s why telepathy can’t happen. We are, each of us, singular and forever alien…”

The amazing thing about people is that this fact doesn’t deter us. We keep trying to share our sensations and feelings with each other. As Virginia Woolf writes in her book Orlando:

For it is a curious fact that though human beings have such imperfect means of communication, that they can only say “good to eat” when they mean “beautiful” and the other way about, they will yet endure ridicule and misunderstanding rather than keep any experience to themselves.

To be known and understood

Maybe we’re just stubborn. Maybe we’re clinging to a shred of hope that science is wrong and someday, instead of just overlapping with pieces of each other, we’ll be able to know what it’s like to experience the sunset the way someone else does. Here’s a passage from a favorite book of mine, Man Walks Into a Room by Nicole Krauss:

“When you’re young, you think it’s going to be solved by love. But it never is. Being close—as close as you can get—to another person only makes clear the impassable distance between you. . . .

“But see, the incredible thing about people is that we forgot,” Ray continued. “Time passes and somehow the hope creeps back and sooner or later someone else comes along and we think this is the one. And the whole thing starts all over again. We got through our lives like that, and either we just accept the lesser relationship—it may not be total understanding, but it’s pretty good—or we keep trying for that perfect union, trying and failing, leaving behind us a trail of broken hearts, our own included. In the end, we die as alone as we were born, having struggled to understand others, to make ourselves understood, but having failed in what we once imagined was possible.”

“People really want that, what did you say, merging souls? Total union?” [Samson]

“Yes. Or at least they think they do. Mostly what they want, I think, is to feel known.”

What do you think? Is the ultimate human goal to feel known and understood? And if that’s the case, is the illusion of feeling known enough to compensate for never truly being able to share one’s experiences with anyone else?

I am not special.

I am human, and there is nothing that makes me special. I am composed of the same atoms as every other thing in existence in this universe. I am no more special than the Jacaranda trees blooming outside my window, than the magpies cawing from their perches on the roof, than the strangers whose eyes catch mine on the street.

And yet I am unique: There is exactly one organization of atoms that is me. There is exactly one set of events that could have given rise to the person I am, because if any other events had occurred, I would not be exactly the same as I am right now. It’s almost so obvious it isn’t worth stating. Almost.

Copernicus and Galileo weren’t special, either

Humans have always held a biased view of their existence. We placed ourselves in a geocentric, Ptolemaic universe, with the stars and planets revolving around us. We were reluctant to abandon our pedestal: When Copernicus proposed a heliocentric model of the universe, supporters of the theory were condemned for heresy. Galileo was lucky: He was just placed under permanent house arrest. Giordano Bruno was burned at the stake. Fast-forward through a few hundred years of debate and you reach today, when it is accepted that our planet does not sit in a privileged location in the universe, but most people are still uncomfortable with the idea that we might just be here by chance.

In a universe of 70 sextillion stars (and that’s just the observable universe, according to a 2003 estimate), it is nearly inevitable that at least one planet would form in the habitable range around a star, and that the right combination of elements would be present for some form of life to develop and evolve enough for beings conscious of their own existence to arise (Argyle, 1977; Ellis & Brundit, 1979; Hoyle & Wickramasinghe, 1999). Our Earth happens to be one such planet.

It’s all physical

I’m going to propose something radical: The universe is a purely physical place. If this is true, it follows that everything in it is also physical, including the mind and mental states. This tends to be a huge point of contention, as the majority of the people in the world adhere to a dualist theory of mind, in which the body is physical but the mind is made of a corporeal substance, unexplainable in physical terms and irreducible to physical states of the brain. (More specifically, most world religions adhere to some form of dualism, e.g., the belief that one has a soul, and the majority of the people in the world adhere to one of these religions.)

There are a number of problems with this approach, most notably the fact that the dualist perspective cannot satisfactorily answer the question of how the mind and body can interact if they have no causal properties in common. Descartes proposed a substance he called animal spirits as a go-between, which merely begged the question (Churchland, 1988); 17th century philosopher Arnold Geulincx suggested the interaction was merely an illusion and that the brain and mind don’t actually affect each other at all (Livingston, 2002). The seemingly plausible suggestion that the mind and brain are connected through a form of energy not yet understood is a logical argument from ignorance. Dualists argue that the mental domain is not lawful, that concepts such as truth, love, and beauty are forever beyond reduction–and yet biologists have found ways to reduce life, which was also once said to be irreducible and magical (Churchland, 1988). Why can’t the mind be a similar case?

Another problem to solve was that of consciousness: If the mind is not made of a special substance, then how does matter organized into a brain give rise to consciousness when matter organized in other ways does not? Various theories suggest that it is exactly this–the particular organization of the brain and the sheer number of connections between neurons–that is responsible for consciousness (see, e.g., Edelman & Tononi, 2000; Hofstadter, 2007). The details of these theories I’ll leave for a later date.

Science says

If you are inclined to believe scientific evidence, the theories that win out reduce the mind to matter. I could easily spend a few thousand more words explaining why this is the case, but I’ll move on for the sake of finishing my explanation of the site name.

In a physical world, people (along with all other living things on Earth) originate from DNA in a purely physical process. You may have heard of the “nature versus nurture” debate: Are we just our genes? Is everything we are predetermined by our DNA, or does the environment in which we grow up and live play some role? I’ve never understood why it’s a debate. Studies of twins have shown that possessing identical DNA will not give rise to identical people (Harris, 2007), which disproves the “we are just our genes” hypothesis. DNA obviously has some affect, and Harris cites evidence that genes account for about 45% of personality, but what makes people and creatures (otherwise known as agents) who or what they are is the interaction of their genes and their environment. Every agent is the sum of everything internal to it and its interactions with everything external to it. Humans are no exception. I am no exception.

Thus the title.

References:
Argyle, E. (1977). Change and the Origin of Life. Origin of Life, 8: 287-298.
Churchland, Paul. (1988). Matter & Consciousness. Cambridge, MA: MIT Press/Bradford Books.
Edelman, G., & Tononi, G. (2000). Consciousness: How matter becomes imagination. Penguin Books, London, England.
Ellis, F., & Brundrit, G. (1979). Life in the Infinite Universe. Royal Astronomical Society, 20: 37-41.
Harris, J. (2007). No Two Alike. W.W. Norton & Co.
Hofstadter, D. (2007). I Am a Strange Loop. Basic Books.
Hoyle, F., & Wickramasinghe, N. (1999). The Universe and Life: Deductions from the Weak Anthropic Principle. Astrophysics and Space Science, 268: 89–102
Livingston, K. Integrating the Sciences of Mind. Chapter 2.