These are organized around the major parts of the course. You dont have to stick so closely to the syllabus, but your project should tie in in a substantial way.
Papert's microworlds were mostly math/science domains, and from the concrete (geometry, physics) parts of those domains. Can you come up with microworld examples from other domains, especially more abstract ones? And/or propose another programming language that would be appropriate for those domains?
Pick some existing educational software (or web-based learning system) and critique from both a Piagetian and a Vygotskian perspective. Its easier to do this if the software is directed at children, but you should be able to handle adult learning software as well. e.g. for the Piagetian perspective, look for sensori-motor --> concrete --> abstract progression.
Perform a Bakhtinian analysis of some class of documents. Look for voices of authors, and/or social languages and genres. Ideally, this would be done as a computation using LSA-like techniques.
Cognitive Science:
Do a speech-act analysis of an office task, like making a travel booking or requesting computer support. Try to anticipate the possible discourses. Include low-level recovery acts for common misunderstandings. It will probably help to have real dialogues with others to work from.
Try to uncover non-obvious metaphors embedded in a user interface. Apart from desktop, folders etc., look for other naturalistic models in the way people talk about the interface. Or, do this for the web. You will need to get some samples of dialogue from the way people talk about the web.
Apply the ideas from the media richness paper to a non-office domain, e.g. to a collaborative design task. Suggest a variety of digital media that could be built into a computer-aided design tool and how "best practices" of use could be encouraged.
Personality, Emotion, Persuasion:
Propose a scheme for analyzing the personality and emotional content for a web site. Relevant factors are language, use of color, image vs. text content and other aspects of design. Can you extract such information by hand from aggressive commercial sites (sites that try hard to promote something)? Alternatively, write a program to do this.
Do the same thing for an interactive agent such as one of the Extempo Characters (several can be downloaded and run on your machine). Do you notice emotional shifts during dialogue with these characters?
Classify the persuasive strategies employed in say: (i) Microsoft Powerpoint Templates, (ii) Several commercial websites. Pay special attention to strategies where the persuasion (and strategy) are not obvious but subliminal.
Program an avatar or physical character (we have one 18 degree-of-freedom SONY dog available for this) for a variety of affective displays. Write a program for sensible transitions between those displays, ideally in response to stimuli from a human interlocutor.
Critique one or more teleconferencing systems (Netmeeting and Mbone tools are freely available) from the point of view of non-verbal communication. How could you improve the breadth and quality of cues sent and received?
Look for social network structures (implicit groups, central or prestigious individuals, similar positions) from some real dataset. We have data for some email communication available for this, and some log data from an ISP. You can also try to mine public newsgroups.
Social networks provide an interesting way to approach Vygotskian and peripheral participation models of learning. Using some data over time (which could be your email interaction within a small group), see if you can find a pattern of "peripheral to central" communication. Could this be made a viable form of assessment in team learning?
Study diffusion of some product over time. You will need the ISP dataset mentioned above, or some similar dataset to work with.
Design and Knowledge Creation:
Perform an ethnographic analysis of some user community for a product you would like to develop. Perform some participatory design cycles by including the users in "envisionments" and simulations of the new product.
Give a breakdown of the tacit and explicit types of knowledge in some domain, e.g. the research group that you work in. How does conversion happen? How could you facilitate it? Are there ways to make more of the tacit knowledge explicit?