[quoting Tripp, 17 May 99.a] (a.) Unreliable technology.
This depends on what technology you use. Unfortunately, today's most used Operating System technology (Microsoft Windows) is not the most reliable one. Furthermore, the capabilities of the Internet are quite over expected; many people dream about streaming video and audio, however a common university Local Area Network can only handle a modest amount of those sessions. These technologies cannot (yet) be used on a university-wide scale.
(b.) Awkward interfaces. Virtually all online learning environments that I have seen have an impoverished discourse model. Essentially they are finite state automata (i.e., they can assume a finite number of states...
In essence, all computer programs can be reduced to finite state automata. When I was a Computer Science. student, we had to read a book by Sudkamp (Languages and Machines) in which there was a theoretical proof that computers based on the Neumann model (central processing unit and a memory with instructions) can be translated into Turing machines. A Turing machine is a finite state machine.
...and they offer limited choices in each state). This is still a Skinnerian model of interaction, however much it is disguised. I would like to see a machine that "knows" me and frames situations based on that knowledge.
The state information would be a model of your knowledge, and there would still be a finite number of choices. Many researchers on the field of Intelligent Tutoring are working on student models, knowledge models and so forth. I think the human thought patterns are too complex and too different from person to person to be put into a single unifying model.
It would also let me ask questions about questions. Not question after question.
Computers already have great problems in communicating through natural human language, let alone that they can do meta-communication. The amount of intellectual capabilities that are required by the computer program you describe are more than what can currently be built. I think that the era in which in becomes possible to build such programs, is also the era in which we build programs that can really "think."
2. The question about optimizing the utility of resources is important, but there seems to be little interest among ed tech people. The work being done in knowledge objects and ontologies is interesting and relevant. There is a professor at my university modeling knowledge objects as "films," essentially multimedia knowledge objects that "know" about themselves and therefore can be queried, combined, shared, linked, etc. This is interesting but contrary to fundamentalist constructivist theology, which is a roadblock to progress in this area.
My Ph.D. research also involves retrieving multimedia learning objects. In the first year of my research I considered using knowledge objects, but if you use knowledge objects (and concept maps or the like) then you easily end up solving Intelligent Tutoring problems. This was not the direction I wanted to take, so I focused on modeling learning objects, i.e., objects of which I don't know what knowledge is in there, but I do know for what educational purpose it was built, for what education level, using what pedagogical principles. A teacher can then formulate a query using these terms and retrieve learning objects that fit his/her description. A lot of work is being done (also by ed tech people) to standardize the descriptions above (called "metadata," see the IEEE P1484 LOMG group, the IMS project, and the Ariadne project). My work builds on this metadata to develop a more formal and mathematic retrieval method.
3. "Computers-in-Education" has been greatly oversold and has probably mostly been a waste of money.
I agree.. I think the only benefit that we can get is organizational benefit, i.e., students are less restricted by time and place demands. However, in many situations the lack of interaction with humans (which is replaced by interaction with computers) may lead to less capable students. It is important that nothing is lost when educational programs are re-engineered into learning environments, and social interaction is often the first thing that is lost. Even when video-conferencing is offered as a replacement, things like gaze direction are lost. Research at our university supplied evidence for the relation between the amount of words like "you," "me," or "her" with the presence of gaze direction. One could conclude that conversations become less personal when they are held through videoconferencing.