Tuesday, December 29, 2009

Omission Bias

Imagine we have two sleeping people. One is dying and needs medicine. The other is healthy. The first dies because a person nearby who is capable of giving her a life-saving drug does not do so. The other dies because another person injects them with poison. Which is more in the wrong, the one who failed to give the medicine, or the one who injected the poison?

Even if, intellectually, you feel both are equally at fault, you very probably have a nagging gut feeling that the injector did something worse than the abstainer.

This is the "omission bias," which is the tendency to consider harmful actions as worse than equally harmful inactions. It's debatable, morally, whether this is a bad bias or not. That is, some think it should not be considered a bias at all.

I want to talk about it in terms of governmental drug approval. Kurzweil suggested in his The Singularity Is Near that people are loathe to approve drugs that might kill you. The way he sees it, (and I'm sympathetic to this view) is that people think that giving someone a drug that will kill 20% and save 80% of terminally ill people should not be put on the market. I see this as omission bias at work. People think that letting people die because they did not get a drug is better than killing someone by giving them a drug.

Accepting the omission bias as a bias to be overcome, however, has some disturbing ethical results. Is there anyone on Earth that has ever died because of your inaction? Let me rephrase that: are there any actions you could have taken in the past that would have saved anyone at all? People who could not afford medicine? Is there anything you can do now that would save the life of anyone in the future? If so, and you believe that omission bias is actually an illegitimate bias, then you're as guilty as a murderer.

Bookmark and Share

Friday, December 25, 2009

I've been meditating for about a year

I'd toyed with the idea of meditating for several years, and read a few good books on Buddhism and meditation, mostly by Thich Nhat Hanh. I highly recommend my favorite book on being happy: Teachings On Love:
If I'm ever feeling angry, this book always soothes me and makes me feel better.

However, the way Hanh recommended meditation never worked for me. It was frustrating and too difficult. I never really got into At the same time I had a growing feeling that meditation is something I should be doing. It's been proven to have several health benefits.

One book made it happen for me. I heard of it by asking my facebook friends for a recommendation. A perceptual psychology professor at Georgia Tech recommended Meditation by Eknath Easwaran. I read it, and loved it. I recommend it highly.

It did two things for me. First, it convinced me that I should be meditating in the first place, and why. Sometimes I get judgmental and angry, and I can't do much about it. I sometimes find myself thinking of things that have made me angry. Sometimes these events happened as long as 20 years ago. I can get myself worked up into a real fury for no good reason at all. Lately I've had a bit of anxiety, too. Not as much as most people, I don't think, but enough to be a problem. I meditate to get control of my own mind. I'm an athiest, so I don't buy into many of the mystical beliefs surrounding meditation (more on that in another blog entry). I just want to be able to stop myself from being miserable, anxious, or angry, whenever I want.

Second, it showed me a way to meditate that I could actually stand doing. Hanh recommends mindfulness meditation, which I could not get to work. Easwaran recommends mantra meditation, which I like better. Below I will describe exactly how I meditate, but first I will report on the results.

In short, not much, yet.

I think I got a lot out of the Meditation book. It is full of good ideas I was a happier person, better able to deal with the crap life throws at you, right after reading it. It's worth reading even if you don't start meditating.

However, in terms of the benefits of meditation, I have not seen them yet. I just talked to someone who has been meditating for 30 years, and he said he did not start to see results for two years. It's really a huge commitment, and not easy.

As I hinted at above, there are several kinds of meditation, and the one that works for me is mantra meditation. I have a string of words that I repeat over and over in my head (not aloud). I read a book by the Dalai Lama and in it was something like "As long as there is suffering, I am here to serve." I started meditating with this mantram (which I believe is singular for mantra). I would sit with my eyes closed for a half an hour, right after my shower in the morning, and say that to myself over and over.

Since then the mantram has evolved to its present form. I didn't like the fact that it was all about helping others and not necessarily being happy about it. So I changed it to "As long as there is suffering, I am here to joyfully serve." I liked this better, but I wanted more of a focus on happiness, so I added to the end of it "As long as there is happiness I am here to rejoice." To add a calming effect, I added to the end of that "Calmly, Peacefully, Rama." "Rama" basically means rejoice. It's there because it's supposed to be a word that has worked for many people for a long time. I was already breaking the rules by making up my own mantram, so I thought I'd throw the philosophy another bone and use "rama" in there.

The trick is to try to think about the mantram and only the mantram for half an hour. This is incredibly difficult, because your mind wanders. If you can control your mind, it won't wander. This is what controlling your mind means. I think of meditation as a workout, training my mental muscles to concentrate on what I want to think about, and not be subject to whatever my subconscious throws at me.

To make it easier, I focus on a different part (usually a word) each time I go through it. So, the whole mantram is "As long as there is suffering, I am here to joyfully serve. As long as there is happiness I am here to rejoice. Calmly, peacefully, rama."

The first half is darker, focusing on the world's problems. The first part I focus on is "as long." I picture the entire Earth, rotting and black, filled with all the terrible hurt and pain that there is. Just to be clear, for each word I go through the whole mantram in my head, focusing on the word or phrase, and its accompanying image, during that go-through.

Next is "suffering." I think of someone or something in pain. A crippled dog, a lonely old person, a woman in Africa being threatened by children with guns, a man who had a stroke, someone in the hospital in chronic pain, etc.

Next is "I." I picture myself as a heroic figure, standing in the darkness, ready to help all of those in need. It's kind of inspiring.

Next is "here." I focus on the word "here" as a nod to mindfulness meditation, another kind of meditation in which you focus on your sensory surroundings and what goes through your mind. I think about where I am, how my body feels, and imagine a connection between me and anybody else in the room. Since I meditate before my wife wakes, usually this is just my dog. To distinguish it from the "here" in the second part of the mantram, I imagine the connection between us made of a spider web, slightly illuminated by moonlight.

Next is "joyfully." It's important not only to help people, but to be happy doing it. I smile while I focus on this word, and imagine a crescent moon above me, to distinguish it from "rejoice" in the second part. Just making yourself smile has been proven to make you feel happier.

Next is "serve." On this I open my eyes to remind me that I'm dealing with the real world.I often look out the window at the sunlight playing on the building across the street.

Next is the happy half of the mantram. I focus on "As long" again, but this time around imagine the whole earth, glowing with love and happiness, representing the good feelings everybody has.

Next is "happiness." I picture someone or something very happy. Usually I picture my pug, Mrs. Wiggles, when I'm rubbing her belly, which is an easy-to-retrieve image of pure joy. Other times I picture my wife laughing.

Next is "I." I picture myself again, as a heroic figure, but this time I picture myself in a sunny environment.

Next is "here." I be mindful of my environment again, and imagine a glowing connection to whomever is in the room. Again, usually it's Mrs. Wiggles.

Next is "rejoice." I smile and try to appreciate all the good in the world.

Next is "calmly." Thinking of being calm really does calm me. To keep track of where I am, I focus on the letter C.

Next is "peacefully." Same thing, focusing on P.

Next is "rama." This time I really focus on the meaning of the sentences.

I call that a cycle. I say the mantram 14 times, each time focusing on a different part. Three of these makes a "supercycle." I keep track of the cycles with my hands. For the first cycle, I have my hands resting on my legs. For the second cycle, I hold my hands together. For the third, I cradle my right hand in my left, with my fingers touching. This is a common meditation hand pose.

I know this sounds complicated, but I need all of these clues to keep track of where I am.

After one supercycle I take a break. My half hour meditation session involves three supercycles and two breaks.

The first break involves thinking about the people in my family and two of my close friends. I imagine my mother, father, wife, self, and my sister's whole family, one by one, and feel love for each of them. When I think of myself, I think of myself as someone in need of love, as opposed to the strong, heroic image I use during the cycles.

The second break is a focus on my body and health.

So each supercycle involves 14 x 3 repetitions of the mantram, which is 42. I have three supercycles, so during a half hour I repeat the mantram 126 times. With the two breaks, it takes one half an hour.

I am committing myself to do it for another year. If I don't see progress by then, I will reconsider.

Bookmark and Share

Tuesday, December 22, 2009

Do you think the world is ultimately a fair place?

One of the things that bothers me about many religions is the belief, often tacit, that the world enforces its own justice. We can see this in modern Christianity (the good go to heaven, the bad to hell) and in Hinduism (the good escape the cycle of reincarnation, the bad are reincarnated as a lesser creature to endure this world's hardships again).

Turns out people in general want to believe that the world is a just place, even without some made-up afterlife. The tragic consequence of this is the "just-world" phenomenon
which is that people, when witnessing an inexplicable injustice, will look for reasons that the victim deserved it.

Please keep this in mind, and have compassion. If the injustice appears in explicable, it just might be. In general, the world is only fair to the extent we make it so.

Merry Christmas.

Bookmark and Share

Saturday, December 12, 2009

Some non-human animals store truth values for propositions

When I tell you that Michael Dukakis is the president of the United States, you will have no problem understanding the statement. If I ask you about it later, you will have no trouble remembering it. You also know that this statement is false. This indicates that we store, in our minds, not only facts, but whether or not they are true. The truth status of a statement is called its "truth value."

I'm going to use the word "proposition" rather than "fact," because "fact" implies truth, and I'm going to be talking about propositions that might not be. I'm going to use "belief value" rather than "truth value" because what I care about is not whether or not it's actually true, but whether the person storing the fact believes it's true.

It's very important for humans to be able to represent belief values, particularly because we communicate with each other. People sometimes tell us things that are false, either because they are attempting to speak truthfully on a false proposition they believe, or they are trying to deceive us.* As such we need to know who to trust and who is unreliable.

Can non-human animals do this? Vervet monkeys (pictured) can. Vervets have calls they make when they spot particular predators (eagle call, leopard call, snake call, etc.). Very cool.  It has been observed that during skirmishes, sometimes a member of a losing troop will sound a leopard call to scatter everyone (reported in Dennett, 1998).  Dennett also describes (can't find the reference) an experimental intervention in which they framed a monkey. Scientists recorded the warning calls of a particular vervet and broadcast the sounds at inappropriate times. Eventually, the other monkeys ignored the calls of that monkey. They learned that he or she was unreliable. I think, at minimum, this means that they are representing the proposition "there is a leopard" along with a tag indicating that it's not true.

I see no reason for any non-communicative creatures to be able to be able to represent belief values. Rather, I think most simply record their perceptual environment into memory, and assume it's veridical. This suggests an interesting experimental prediction: that non-communicative animals cannot learn to ignore an illusion. They might learn to treat it differently, but I predict that they would continue to believe in the existence of what the illusions represent no matter how much you train them.

I would like to think that communication is a feature only of very intelligent animals, but I'm afraid it's just not true. Ants communicate, and even the cells of your body communicate with each other. What I would really like to know is if the belief value of a proposition is stored with that proposition, or if the belief values are stored separatrely, indexed to the propositions. This has implications for cognitive modeling.

* The movie "True Lies" is meant to sound like a contradiction, but it isn't, really. A true lie would be when someone believes a false thing, x, and then lies to you by stating not x. For example, Jack believes in the lunar effect even though there isn't one, but then lies to me by saying it doesn't exist, perhaps to impress me because he knows I'm a scientist. Jack stating that there is no moon effect is a true lie.


Dennett, D. C. (1998) Brainchildren: Essays on Designing Minds. MIT Press. p290

Bookmark and Share

Friday, December 11, 2009

A bit about what I'm interested in most, as a scientist

Visual instantiation is a cognitive process that generates visual and spatial properties. 
Humans visually instantiate when reading, designing, planning, dreaming, problem solving, and creating physical art. The variety of inputs is great. The stimulus that triggers the instantiation could be verbal, for example when one is reading a description of a scene, or when someone asks you to picture something. The stimulus could be internally-generated, as it is when imagining what would happen if a bookshelf fell over. What makes all of these examples similar is that they all create visual and spatial properties.

Imagination is a related process. Broadly speaking, imagination is self-generated input for the perceptual system. It is not limited to the visual modality, as one can imagine how something might taste, and one can imagine the sound of Vincent Price's voice. In visual imagination, evidence suggests that the mind creates patterns of neural activation that is spatially organized in the visual cortex, in the same place as the pattern made during actual perception (Tootell, Silvermanm, Switkes, & Devalois, 1982). In computational terms, each neuron represents a `bit' in a spatially-orgainzed `bitmap.' Kosslyn (Kosslyn, 1994) refers to this kind of image as ``depictive.'' Depictive images represent only colors (including hue, saturation, brightness) at certain points. The generation of a bit-map, or depictive, image is called ``rendering.'' It means going from some structured representation to a bit-map. This structured representation is the `descriptive' image (Kosslyn, 1994). How one gets from a descriptive to a depictive image is something I'm interested in, but I'm even more interested in the generation of the descriptive image in the first place. 

As such, I use the term ``visual instantiation,'' which deals with both the generation of the descriptive image and the depictive rendering. 


Kosslyn, S. (1994). Image and Brain. MIT Press.

RB Tootell, MS Silverman, E Switkes, and RL De Valois (1982).

Deoxyglucose analysis of retinotopic organization in primate striate cortex.

Science 26 Vol. 218. no. 4575, pp. 902 - 904

Bookmark and Share

Thursday, December 10, 2009

Poster I made for advertising my cognitive science institute

I'm on the graduate committee in my department. We wanted to make a recruitment poster and I volunteered to make it. I like the design, although we were forced to put the logos on the bottom right. I think the design and effect is better without them.

Bookmark and Share

Friday, December 04, 2009

Project idea: two neural nets, one data set

Here's a project idea.

Train two neural networks on the same data, but with different initial random weights. They will come out differently, but probably equally good at whatever classification task they are learning to do.

Come up with a way to determine that they were trained on the same data.

I expect that this finding will shed some light on how this might eventually work in brains. Neurally, your belief that 'a' is th first letter of 'apple' probably looks a lot different from mine.

I'm not a neural net scientist, so I'll probably never run this study. If you do, let me know.

Bookmark and Share