Title: “A Computational Algorithm for Sustaining ‘Spatialised’ Communication”

PhD Research Summary – Mauro Cherubini

Communication has suffered from the clearing of spatial references since the invention of the telephone or the radio. Nevertheless, spatial informations are used all the time by the agents-in-relation to reconstitute the context of the messages exchange. The effect is an economy of the whole interchange process. The aim of this research is to understand by which mechanisms this happen. Particularly, the focus will be on the shared knowledge construction. Subsequently, and as a proof technique, a formalisation attempt will frame these possibly uncovered processes into an algorithm. The goal will be to replicate, to some extent, human reasoning and sustain communication exchange when applied to real time interaction.

An implementation of this algorithm will be applied, tested and refined in a geographical messaging system developed by an English firm called Proboscis (see http://urbantapestries.net).

After a meeting with my advisor, Pierre, I was re-thinking the idea of defining a strategy for grasping the co-referenced actions. These assume a very central role in our theory because we equate them with grounding at a knowledge level.

In his old work on the Bootnap project, Pierre, was more focusing at the utterance level of analysis and therefore he was not using the acknowledgment by action consistently to determine grounding at a more abstract level.

Consequently, the interesting edge of this research may be the definition of a methodology able to detect when a person has a shared understanding (abstract) at a certain moment in time. This methodology, or coding schema, should focus on observable elements of the interaction, actions, speech, writings, etc., and should put them into a framework for the detection of Mutual Modelling acts or Perspective-taking acts.

De distinction we tried to sketch with Pierre, was a bit simplistic: we discriminate the two concepts saying that Mutual Modelling is a kind of diagnosis the agent does of the undergoing communication, whereas the perspective taking is more the adoption of another agent’s point of view.

This coding schema, could, in turn, constitute the core of the computational algorithm used to automate the process of computationally sustain the conversation.

1. Designing and testing collaborative technology-enhanced furniture, such as noise-sesntive tables. Looking for a creative techie having done his PhD or willing to do a PhD in the field of roomware, tangibles and augmented reality. Duration: 2-3 years. Start: ASAP

2. Design a high level language for modelling CSCL scripts. Scripts are generic scenario that structure teamwork with roles, phases, etc. Looking for a computer science PhD student with an interest for abstraction. Duration: 3 years. Start: ASAP

3 & 4. Designing and running experiments on mutual modelling, i.e. the representation that team members have of each other. For this NSF project, we are looking for two cognitive science researchers, a postdoc and a doctoral student. Duration : 3 years. Start: spring 2005

Interested? Send me your CV! Pierre.Dillenbourg@epfl.ch

A reshaped definition of Mutual Modelling from Nico:

Mutual Modeling: process by which an individual (i) picks up facts he/she estimates to be MANIFEST and RELEVANT in his/her cognitive environment (with regard to the partners’ intents), (ii) builds a representation (more or less formal) about the partners’ intents.

Found a nice NMEA parser written in Python. As a plus it allows to visualise Maps downloaded from Terraserver. As an interesting side-link, I discovered that pygps requires image (bitmap) maps which are georeferenced. The author mostly use terraserver maps, but any map with a world file will work. He suggests to go to http://www.acme.com/mapper/ to create such kind of maps.

We spent the day in Morges, enjoying the sunshine and the lake side. The city was quite alive and the landscape just astonishing.

I started looking into this curious book titled: “Change the world for a fiver -50 simple actions to change the world and make you feel good“. Apparently in Great Britain, this book was realised to launch a movement called We Are What We Do, with the underlying idea that if anybody start doing -the same- very simple and little thing, then the final effect would be tremendous.

I support -strongly- this idea and I will report time to time the ideas that I found very interesting. I will start today with the concept of Trashware:

Trashware often means “badly designed or useless software,”  but in recent years the term has came to indicate something else which could have an important impact on computer users. I use trashware to refer to trashed hardware. We all see perfectly good working computers dumped as soon as the latest and cutest software slows them down. But trashware is not trash, and a number of groups are working to save those “obsolete” computers from the landfill and using free software to make them useful again for fun, profit, or service.

There are lots of resources on the net on trashware. Here an initial list of links.

Krauss, R. M. and Fussell, S. R. (1991). Perspective-taking in communication: Representations of others’ knowledge in reference. Social Cognition, (9):2–24. Available from: http://www-2.cs.cmu.edu/~sfussell/pubs/pubs.shtml.

—————

To communicate people need extensive assumptions about what others know, want and believe. This is the core assumption of this paper, in which the authors want to spot some light on how people take the role of the other while communicating. These assumption are necessarily tentative and probabilistic and therefore, agents may come to different conclusions about what is “mutually known”.

The authors report the fact that little research has involved the mechanisms that allow shared knowledge, beliefs, or perspectives to be inferred. Also, they stress out the fact that as with other forms of social reasoning, people may utilise a variety of knowledge structures (e.g., schemata, stereotypes, inference heuristics) to estimate what others know. These structures may facilitate the inference process but, in the same way, they can introduce sistematic errors or biases.

Found an interesting site devoted to satellite television in europe. It offers technical informations on satellite transmissions, bouquets, transponders, etc. There is also an interesting section on web tv broadcasts.

[www.lyngsat.com]

The view from the Hercules Temple, Cori, Italy.