My first thought was, no, Watson did not actually learn anything. It has been programmed to perform, but it can only perform within the parameters the programmers established. To me, the strongest element of learning is the ability to take new information from one situation and apply it to another. However, after looking in the dictionary, I realize my definition may be off.
In fact, Merriam-Webster’s online dictionary lists the first definition of the verb “to learn” as follows: “to gain knowledge or understanding of or skill in by study, instruction, or experience” (Learn, 2011). This definition doesn’t include the ability to transfer the information to new a context, like I had original thought. It also states that one learns by instruction, which is conceivably equivalent to programming, in the case of Watson. So, perhaps Watson has learned the information it relies upon to answer trivia questions.
Beyond learning, there are the concepts of knowledge and understanding. Wiggins and McTighe (2005), in Understanding by Design, make it pretty clear that there is a difference between knowledge and understanding. Wiggins and McTighe state, “An understanding is a mental construct, an abstraction made by the human mind to make sense of many distinct pieces of knowledge” (p. 37). From this description, we see that understanding is on a higher level of cognition—a level that is attained by applying “distinct pieces of knowledge.” Therefore, knowledge is a building block that leads to understanding. This concept is also apparent in Bloom’s Taxonomy, which places Comprehension (understanding) above Knowledge (Page, 2010, p. 56).
One can argue that Watson learned a variety of facts, which is the knowledge Watson relies on to answer the trivia questions on “Jeopardy!”, but does Watson understand? Again I’ll rely on Wiggins and McTighe, who state, “Understanding thus involves meeting a challenge for thought. We encounter a mental problem, and experience with puzzling or no meaning. We use judgment to draw upon our repertoire of skill and knowledge to solve it” (p. 39). By this definition, one could even argue that Watson understands the game show prompts, at least some of the time, as evidenced by the computer’s ability to sort through its “repertoire of skill and knowledge” to answer the prompt correctly. Of course the computer can make mistakes, but, then again, so do humans and animals, so why not computers.
However, as we delve deeper into the layers of understanding, it seems less and less likely that a computer, even one as well programmed as Watson, can be said to truly understand. Wiggins and McTighe present six facets of mature understanding:
When we truly understand, we,
- Can explain
- Can interpret
- Can apply
- Have perspective
- Can empathize
- Have self-knowledge (p. 85)
While it may be silly to look at learning, knowledge, and understanding as it applies to a computer, I believe it to have been a valuable exercise nonetheless. As humans, we perhaps take for granted the fine differences between these three concepts—in fact, we often use these terms interchangeably—but, by applying them to an inanimate object, we have to really examine our definitions of these concepts and determine how they might be applied to something outside our normal frame of reference. For that matter, this was an exercise in understanding the concepts of learning, knowledge, and understanding by transferring what I’ve learned in one context (the typical teacher-student relationship) to that of a different context (the programmer-computer relationship).
If you’d like to read more about Watson’s big match, here is an article from the New York Times, and one from the Washington Post.
[LIBR250 Prompt: What does it mean to learn? Include thoughts and definitions regarding knowing and understanding?]
References
Learn. (2011). In Merriam-Webster online. Retrieved from http://www.merriam-webster.com/dictionary/learning?show=10&t=1297730018
Page, B. (2010). 12 things teachers must know about learning. Education Digest, 75(8), 54-56.
Wiggins, G. & McTighe, J. (2005). Understanding by design (2nd ed.). Upper Saddle River, NJ: Pearson.
10 comments:
Interesting! I find AIs fascinating, as a devoted fan of science fiction. They certainly trigger questions about what it means to know, to learn, and to be human.
I would even argue that perhaps Watson can't explain, but can only interpret and apply. Then again, it probably isn't being asked to explain, so we will likely never know.
One question we could ask of the computer to determine whether it can learn is this: when it makes a mistake (as you said it does sometimes), what does it do? Can it add the new information to its own base of knowledge? Can it explain its own thinking processes that led to the wrong answer and correct them? Could it (as you suggested) apply its knowledge in a non-Jeopardy-related setting?
If so, then the game is afoot!
You certainly make a good point about Watson not being able to explain. We watched Jeopardy! tonight and were wondering if Watson learned from its mistakes. For example, one category required a decade answer rather than a specific year, and at first Watson just couldn't get it, but then it got one right. Interesting.
Nice way to bring a current issue into the conversation as an example. This also skirts along issues of cognitivism and constructivism as philosophical areas. We are pretty firmly rooted in cognitivism in our educational and learning paradigms as they currently exist (even when we say we are constructivist) - the individual, how does he/she think? We rely on the individual - what is the prior knowledge and how do we scaffold that rather than what are the sociocultural dimensions of learning? I don't have any easy answers to that. But in thinking about what Watson knows and understands you have established the basis for how we often compare a brain to a computer or vice versa. So all of the sudden I have to ask - what is the sociocultural implications of learning? Or is it all individual?
Based on my study Vygotsky and Zone of Proximal Development for our wiki assignment, I'd have to say that learning definitely involves a social component. Learning is bigger than just the individual experience. You may be interested in this article about one woman's experience with moving to the US from Brazil and learning in cyberspace.
Conceicao, S. (2002, Winter). The Sociocultural Implications of Learning and Teaching in Cyberspace. New Directions for Adult and Continuing Education, 96. Retrieved from http://coe.nau.edu/part_time_fr/principle10-article3.pdf
I thought your entries on the wiki were really interesting - particularly in light of this assignment and the difference with threaded discussions. I had intuited or experienced some of that as an online instructor - hence the playing around with format of conversation. I find that blog groups work as well as the members of the group - there is a lot of individual responsibility to the group and it takes time to develop. By commenting back you are setting a good example for your group on how to build that conversation. I hope people check back or subscribe to comments via email. One of the flaws with the blog is it exists outside of Angel and you have to make an effort to visit.
**This comment is actually from Kathleen. For some reason she couldn't post a comment, so she emailed it to me.**
Hi Kimberly,
I appreciated your use of Watson and AI as a point of reference for considering what it means to learn versus understand. Considering Watson’s performance in light of Wiggins and McTighe’s definition of what it means to understand, it is clear, as you say, that Watson may have demonstrated his ability to learn, but he did not show understanding. For it is one thing to be able to explain, apply, or interpret, but something else entirely to demonstrate an ability to empathize, show perspective or demonstrate self-knowledge--and Watson will probably never get there. (And, honestly, I would like someone to explain to me how Watson is able to interpret or apply. This is mind-blowing to me!!)
All of this makes me think about the drive towards standardized testing and the quest for accountability that has driven our educational system for the last ten+ years since the passage of No Child Left Behind. I certainly understand the need for evidence that children are learning and teachers teaching. However, I think we have taken the path of least resistance here, and most states have adopted standardized tests that largely assess a student’s ability to demonstrate learning on the most basic level. Why? Well, we perceive these kinds of tests to be more “objective” and, to be perfectly frank, they’re easier to evaluate--just run ‘em through a machine, and you’ve got your results. Assessing a child’s ability to empathize, show perspective, demonstrate self-knowledge, interpret or apply is far more complex and labor intensive--it involves seeing the child in a far more multi-faceted way.
All of this has huge implications for teaching and learning--a teacher who worries that her job is on the line if her students’ test scores are not continually on the rise will likely begin teaching to the test. Instruction becomes more robotic, and learning becomes a trivial pursuit. In this kind of environment, it takes great courage and confidence for a teacher to buck these trends. But some resistance is in order--our children are so much more than robots.
Technical difficulties resolved!!! I'm in--I'm no longer shut-out from the blog-o-sphere. Phew!!
Actually, your post and the discussion comments have brought to mind the PBS video that Mary Ann linked us to. What is literacy? What does it mean to be a literate and educated person? In the 21st century, that definition is changing. Will kids some day need to simply know where to find information rather than repeat it back to us? The very nature of librarianship follows these lines: we don't have to know the contents of everything found in a library--we just need to know where to find it.
Watson has facts stored in him that he can access, but does that make him "literate" or "educated"? As Kathleen said, kids are not robots. Some people can memorize better than others, and computers can certainly do a better job than any human, as evidenced in the Jeopardy! challenge.
I was looking into this and found a cool youtube video about the challenged IBM faced in creating this computer. Watson had to be able to understand a sophisticated level of semantics to be able to be successful as a contestant on the show. Here's the link: http://www.youtube.com/watch?v=FC3IryWr4c8
I have to admit that there's this part of me that's a little nervous about computers taking over the world now... Too many sci-fi books!
Kathleen, you make a really good point about standardized tests testing only the most basic levels of learning. I'm sure Watson could pass a standardized test with flying colors. In addition to this class, I'm also taking the Research Methods class that focuses on Evaluating Programs and Services. Like you said, standardized tests tend to be more "objective" and they're certainly easier to synthesize the results because a machine can do it for you. But, if the results don't answer the ultimate question of whether students are really learning, then the results--however objective and easy to analyze--aren't very useful. The evaluation tool needs to support the end goal.
Casandria, thanks for sharing the video, and for reminding me about the other video Prof. Harlan linked to in the discussion forum. I'm going to check that out today!
I totally agree, it's a little scary what computers can do these days. For now, I feel alright considering computers still need humans to program them. Watson is no doubt impressive, but really, it's the programmers behind the computer that are really amazing. Now if, in the future, computers are able to some how "reproduce"--build themselves without programing from humans--then I'll REALLY be concerned! :)
Post a Comment