Welcome back! So I've managed to break my emacs. Oh and my macintosh. All in the same week. I'm hoping you don't have to actually have any technical aptitude to get a degree in comp sci... None the less, I've acomplished it and am paying dearly as I cannot figure out how I managed to get the delete key to completely change operations ONLY in emacs, well I have some idea but I can't figure out how to get it to go back. I suppose this may actually involve reading about emacs or using the google. *sigh* Well, how bout I work on that and you feel free to read on. Works for me! Alright, ready set go!
Readings for the week:
The Atomic Components of Thought by John R. Anderson and Christian Lebiere
This week was spent reading the chapter about Learning as well as the tutorials for ACT-R 6.0. Well.. in actuality this week was spent pleading with my computer to work with me as opposed to against me. But, aside from the necessary evils which devoured a great deal of time, I did get a chance to read the Learning chapter as well as a portion of their discussion on "choices" as well as get through three of the tutorial models. I will discuss the tutorials later, for now, to keep myself a bit organized let's stick to the book shall we? The thing I find most interesting and MOST difficult about this all is trying to decide what level should be doing what. The ACT-R architecture does learn in some senses but not in others. It can learn which procedures to favour, it can combine procedures in some senses to act faster, but it can only do this formulaically and if the original data is input in a certain way that makes it possible. I still find myself having to convince myself of things, I am never sure right away if it truly is a limitation of the architecture or merely my way of thinking about it. But the idea of generalizations or abstractions... the idea of planning.. these are all things that have to be done by me, which I can accept but I am still currently deciding if I think this is wise. Should these things be built in, automatic? If so, how? Maybe I'm just being highly demanding or giving too much credit to how we think, maybe we too build up to planning and it is actually a "program" or set of procedures we have developed. That still doesn't explain how it's developed but it isn't basic, it doesn't start there.
One other thing that struck me while I was reading was how to account for individual strengths and weaknesses. Why and how are some people better overall at certain tasks than others? I was having crazy dreams and thoughts about different portions of the brain using different styles of architectures to make decisions or solve problems and the winner is the fastest or more strongly weighted portion of the brain. If you think one way it's because this portion of your brain is faster. But that would mean potentially all portions of the brain fire for every task which I don't believe to be true. Some things have to do with how the brain stores the information. One of the examples in the text deals with categorization where people were told to categorize by one rule and then the rule changed and they observed how long it took the person to adapt. They found that adults were able to deal with changes within a trait easier than trans-trait changes. For example, if the original category rule was that all red objects were 1 and all green objects were 2 and then it reversed, the adults were fairly quick, but if the rule changed to all circles were 1 and all squares were 2, younger children adapted quicker and easier. This was linked to a way the productions were formed, which begs the question why one group is prone to form one set of production rules and the other does not. But this all boils down to how the brain transforms stimuli into productions and data, which is the same problem we encountered last week. We don't know yet. We just assume it's given.
And then this all brings us to the self-modifying architecture. One of the things you do as an experimenter is set values for certain variables and equations for learning. You choose how quickly the architecture forgets or how important certain details are or how strongly certain ideas are linked. The thing is, humans are able to change their own values for these things. When you are in an unstable environment you are less likely to put emphasis on past experience, knowing it may not help, so you make yourself "more forgetful" in a sense or put less weight on previous experience. If you are doing a task you feel comfortable with, or feel you are in a stable environment where things remain constant, previous experience is invaluable and you take this into consideration. Unfortunately, I have been unable to determine if there is a way for ACT-R to modify these parameters during run-time itself. It currently makes sense to me that this would be a valuable trait of the architecture and as such I am hoping to look into if this is possible.
But alas I babble! I shall cut myself short and venture to a new subject. I end with a cheesy quote I like despite it's cheesiness. It's been on the brain for a bit: Be careful of your thoughts for your thoughts become your word. Be careful of your words for your words become your actions. Be careful of your actions for your actions become your habits. Be careful of your habits for your habits become your character. Be careful of your character for your character becomes your destiny.
So that is a snippet of what I've been learning! The language Act-R speaks! I can't say that the tutorials will be very exciting to you and so I will not dwell incessantly on this. I will just briefly mention how heartening it is to finally get the tutorial example to "tell you how many x's are on a screen." I think everyone in my lab found out when that one finally worked. It's interesting though, because it's like learning a new computer language but with an entirely different set of constraints. Just like I can understand certain aspects of C as a bi-product of a limitation of assembly language or as the result of the memory lay out on a computer so can I begin to see the reasoning behind certain choices in the ACT-R architecture reflecting certain aspects of the brain and human thought capabilities. It's very frustrating to get out of the elegant optimal code mentality of writing and adapt to the fact that humans require some redundancy or certain steps which seem easy enough to combine need to be kept separate. And sometimes you have to snap out of thinking an old way, because these new "variables" don't do what they usually do. And here's some new keywords... and that about covers it. You think you got it? good here's your project for this chapter of the tutorial and GO!
Summer gives a new face to the university. Not only are the lilacs in bloom in front of the building, and the grass finally green again, but the people inside are different, their purpose changed, and the entire place has a new energy. You start out kinda shy because you're used to the same handful of people who are in all of your classes are all gone for the summer, and everyone here seems to know exactly what they're doing except for you. You feel like this incompetent waste of space that bumbles about and gets in everyone's way. But then the end of the day hits and the room which was once silent except for the muttering of fingers upon keys fluttering in the background and all of a sudden everyone is talking about the movie "Lucky Number Slevin" or lunch hits and someone catches you when you're slinking off and demands you try the new pizza with them. And you start to fit in. And this smaller community gains a routine. Certain labs get associations, reputations if you will, and you only have to which which doors someone's prox card opens to get an idea of who they are.
This thursday was the introduction coffee klatch to the summer students. Coffee and doughnuts and chilling out in the Heritage Lounge in Athabasca hall with the professors and the grad students. At this event I even ran into Eleni Stroulia who was the woman who told me about CDMP awards and introduced me to Renee in a sense. She was there with HER CDMP student, Lenore Brown, and we took some pictures for her website (see links page in future for link to her website). Eleni is a wonderful lively lady who is part of WinCS, which is the women in CS group on campus, and decided to encourage some of the girls working on campus this summer to get to know each other. As a result I've been invited to attend a "symposium" (which in Greek, her native tongue means "to drink together".. I can't believe I almost wrote which in geek.. I'm such a loser ) at the faculty club on campus on June 2 (read: see future week updates for details) for chatting and delighting in one anothers company. I can abide by this and am much excited to go!
On friday I attended another pizza talk. Our guest lecturer was Eric Hansen and he was discussing some new algorithms for "exploiting graph locality in external-memory graph search". Now boy oh boy did I not know a think about htis before. But it was interesting none the less, more an exercise in exercise, as I tried to chase after the discussion that was a bit over my head.
In other news: Happy Mothers Day! That's about it for this week! I'll see y'all later. Cheers! Leah