pedagocical agents
TRANSCRIPT
Pedagogical agentsPedagogical agents TThe experiencehe experience of Consorzio FOR.COM. of Consorzio FOR.COM.
Mikail FeituriMikail FeituriICT managerICT manager
Consorzio FOR.COM.Consorzio FOR.COM.
Rome, 23 October 2008Rome, 23 October 2008
Intelligent agentIntelligent agent
software elements being responsible for carrying out given tasks by means of artificial intelligence techniques
Conceptually, the agents implement a metaphor being common to the typical way of operating in the market: visiting a place, using a service (possibly following a
negotiation) then moving elsewhere. After the agent has gathered the results
wished, it goes back to the user.
Pedagogical agentPedagogical agent
Definition: particular type of intelligent agent actual virtual tutor accompanying the student of the
educational system during the learning process
Features: always visible to the user within the educational milieu human (or humanoid) forms Interacts with the user both verbally and non verbally It moves and interacts directly with the learning milieu
and within the milieu itself.
Parmenide projectParmenide project
Goals: Two pilot applications for training of
operators employed in the transport sector in the anti firing security field.
Features of the applications: Innovative assessment tool Extremely stimulating scenarios for the
students The virtual tutor simulates a teacher who
submits an exam to a teacher
Platform
BEHIND THE PILOT APPLICATION BEHIND THE PILOT APPLICATION
The pilot application starts on choosing randomly an important question among those available.
Our expert in anti firing security has defined which questions have to be considered as important.
Another parameter, which has been considered, is the difficulty of the question.
The system works with 3 Fuzzy Logic inference system (FIS).
Fuzzy Logic, with its linguistic rules, simulates human behaviour. In fact, it translates human behaviour based on natural language syntax in an artificial language suitable for computers.
FUZZY LOGICFUZZY LOGIC
First Fuzzy inference engineFirst Fuzzy inference engine
FIS 1
Importance
Difficulty
Fastness
Correct / Incorrect
Knowledge depth
It defines a learning path for the It defines a learning path for the studentstudent
The knowledge depth is the degree of user knowledge about the topic
Depending on the quality of the user answer, the system provides again another important question or any other question.
The system behaves like a normal teacher
In the pilot application the minimum question numbers is 3 and the maximum is 5
Knowledge depth Knowledge depth
It provides the score carried out by a user when he / she It provides the score carried out by a user when he / she answered a question. answered a question.
Second Fuzzy inference engineSecond Fuzzy inference engine
FIS 2
Importance
Difficulty
Fastness
Correct / Incorrect
Score
It defines the verbal and non verbal tutor behaviour It defines the verbal and non verbal tutor behaviour
FIS 3
Cumulative score
Knowledge Depth(if answer is right)
Verbal and non verbal behaviour (facial expressions)
Score(if answer is wrong)
Third Fuzzy inference engineThird Fuzzy inference engine
More than 100 verbal feedback are stored in the database.
This messages are classified from very negative to very positive.
The tutor decides which one to supply from the third fuzzy engine output.
Verbal behaviourVerbal behaviour
The tutor is able to provide 11 different facial expressions
The tutor puts on a neutral expression when he reads the questions and she provides the didactic pills.
The tutor decides which one to supply from the third fuzzy engine output.
NonNon Verbal behaviourVerbal behaviour
We tried to avoid virtual tutor behaviours which can be classified as unstable.
For this aim, we considered the user performance carried out in all the questions and not just in the last question answered.
On doing this, we tried to simulate the behaviour of a normal teacher who submits an exam to a student.
NonNon Hysterical behaviourHysterical behaviour
Remarks and improvementsRemarks and improvements
The number of questions is very limited because this is a pilot application for testing new didactic methods.
Only multiple choice questionnaire for each scenario has been used because of the particularity of the didactic topic
Among other sectors, more complex and various scenarios could be used.
Looking ahead: TLooking ahead: T2 2 projectproject
T2 adapts and transfers the pedagogical and didactic model developed in PARMENIDE in the field of microfinance
The aim is to apply the “PARMENIDE model” to a comprehensive and already produced E-course
Looking ahead: COACH BOTLooking ahead: COACH BOT projectproject
COACH BOT is a pilot project that aims essentially to develop an intelligent tutor
Like a real tutor, the pedagogical agent will provide help, suggestions on the lessons, in-depth information, ...
For this, the development should be focus on the agent’s dialogue capacity with the student
The artificial intelligent techniques to be used will be probably rather different from the ones developed for PARMENIDE.
Thank you!Thank you!
Mikail FeituriMikail FeituriFOR.COM. Interuniversity ConsortiumRome – [email protected]+39 06 37725542