1. Use of Artificial Intelligent Agents, Multimodal Integration, and Theory of Evidence to Design Software to aid First Grade Level Math Education By: John Donath Bethel Lynn

2. Introduction Currently Existing Software in Elementary Math Education Application of Agents to Existing Software Communication with Students

3. Beth’s Experiment 25 question Math quiz to students 5 question online quiz to students

4. Beth’s experiment Results Students enjoyed the online quiz more Students learned better from questions on there level

5. Extensions of Beth’s Experiment Use AI agents to customize questions and tutorials to students Use Multimodal integration to help communicate with students without overhead of teaching them computers

6. Available Technology – AI Agents in Education Colleges and High Schools have used the following agents: Digital TA Digital Tutor

7. Available Technology – Voice and Ink Recognition VoiceXML for communicating voice data between browsers InkXML for communication pen data between browsers

8. Integration of Multimodal Inputs improves quality of communication Use Voice and Ink XML technologies to program recognition of various inputs and communication with students

9. Integration of Voice and Ink Inputs Integration of Inputs from Voice and Ink recognition technologies can be achieved by Dempster-Shafer Theory of Evidence

10. Dempster- Shafer Theory of Evidence Each proposition is assigned an interval [belief, plausibility] between 0 and 1. Belief is the minimum probability that the proposition is true Plausibility is the sum of all the probabilities that do not carry evidence against the probability

11. Dampster-Shafer Theory of Evidence Dempster’s rule of Combination Demster’s Rule – in generalDemster’s Rule – in general

12. Example of Theory Of Evidence Suppose we have a set of four words to be recognized: chat (c),finger (f), mile (m), seven (s). Evidence supports hypothesis that the following words were given by a user: chat, finger, mile {c, f, m}. with a belief of 0.6. All other sets therefore have a belief of 0.4

13. Example of Theory Of Evidence Suppose new evidence is found that the user gave the words: chat, finger, seven {c, f, s} with a belief of 0.7. Meaning that the belief of all other sets is 0.4

14. Example of Theory Of Evidence The combined belief is calculated as: M1M2M3 M1{c,f,m}=0.6M2{c,f,s}=0.7M3{c,f}=0.42 M1(Q)=0.4M2{c,f,s}=0.7M3{c,f,s}=0.28 M1{c,f,m}=0.6M2(Q)=0.3M3{c,f,m}=0.18 M1(Q)=0.4M2(Q)=0.3M3(Q)=0.12

15. Example of Theory Of Evidence Suppose evidence indicates that mile alone was the word given by the speaker, with a belief, m4(m)=0.8.

16. Example of Theory Of Evidence Combined beliefs are calculated: M3M4M5 M3{c,f}=0.42M4{m}=0.8M5{}=0.336 M3(Q)=0.12M4{m}=0.8M5{m}=0.096 M3{c,f}=0.42M4(Q)=0.2M5{c,f}=0.084 M3(Q)=0.12M4(Q)=0.2M5(Q)=0.024 M3{c,f,s}=0.28M4{m}=0.8M5{}=0.224 M3{c,f,m}=0.18M4{m}=0.8M5{m}=0.144 M3{c,f,s}=0.28M4(Q)=0.2M5{c,f,s}=0.056 M3{c,f}=0.18M4(Q)=0.2M5{c,f,m}=0.036

17. Example of Theory Of Evidence New Denominator = 1 – 0.336 – 0.224 New Beliefs are: M5{M} = 0.545M5{c,F}=0.191 M5{} = 0.56*M5{C,F,H}=0.127 M5{C,F,M}=0.082M5(Q) = 0.055 * No denominator used to obtain this value

18. Probability Model Views all sets as independent of each other. A set that does not provide supporting evidence is viewed as providing negating evidence Assume that words and sentences are independently recognized, without correlation between them

19. Conclusion Use of Multimodal communication and AI agents can help students improve Eliminate overhead of teaching young students to interact with a computer Customized education and monitoring of each student’s progress