"... In PAGE 8: ... o The Language Learning: Classroom Project: Four students in a Language Tech- nologies course at CMU used the Pseudo Tutor technology to each build two prototype Pseudo Tutors related to language learning. In order to estimate the development time to instructional time ratio, we asked the authors on each project, after they had completed a set of Pseudo Tutors, to estimate the time spent on design and development tasks and the expected instructional time of the resulting Pseudo Tutors (see Table1 ). Design time is the amount of time spent selecting and researching problems, and structuring those problems on paper.... In PAGE 11: ... Pseudo Tutor authoring opens the door to new developers who have limited programming skills. While the Pseudo Tutor development time estimates in Table1 compare favorably to past estimates for intelligent tutor development, they must be considered with caution. Not only are the these estimates rough, there are differences in the quality of the tutors produced where most Pseudo Tutors to date have been ready for initial lab testing (alpha versions) and past Cognitive tutors have been ready for extensive classroom use (beta+ versions).... ..."

"... In PAGE 8: ... o The Language Learning: Classroom Project: Four students in a Language Tech- nologies course at CMU used the Pseudo Tutor technology to each build two prototype Pseudo Tutors related to language learning. In order to estimate the development time to instructional time ratio, we asked the authors on each project, after they had completed a set of Pseudo Tutors, to estimate the time spent on design and development tasks and the expected instructional time of the resulting Pseudo Tutors (see Table1 ). Design time is the amount of time spent selecting and researching problems, and structuring those problems on paper.... In PAGE 11: ... Pseudo Tutor authoring opens the door to new developers who have limited programming skills. While the Pseudo Tutor development time estimates in Table1 compare favorably to past estimates for intelligent tutor development, they must be considered with caution. Not only are the these estimates rough, there are differences in the quality of the tutors produced where most Pseudo Tutors to date have been ready for initial lab testing (alpha versions) and past Cognitive tutors have been ready for extensive classroom use (beta+ versions).... ..."

"... In PAGE 8: ... Information was directed from the tutor to the student, just as we saw in the Expert Tutoring condition. As shown in Table3 , our data set demonstrates that a peer, with a lower average number of interactions, intervenes slightly less than a teacher. The average number of interactions between the teacher and the student was 10.... ..."

"... In PAGE 12: ... Group 1: Distribution of Off-Topic turns OT-S 80% OT-A 13% OT-T 7% Group 4: Distribution of Off-Topic turns OT-S 84% OT-A 9% OT-T 7% Figure 6: Distribution of Off-Topic Turns (Groups 1 and 4) The results also highlight an interesting aspect in the distribution of OT turns by the tutors. A be- tween tutor comparison ( Table5 ) revealed that both tutors produced mainly OT-S and OT-A for the respective purposes of development of social relations within each group, and class manage-... ..."

"... In PAGE 4: ... We present quantitative and qualitative results to support this. Table4 presents all our case studies showing the type of decision problem, the number of people involved and the approach to making a decision. The case study ID is indicated with an A, B or C to identify the company, ... ..."

"... In PAGE 2: ... Our Earth observing sensorweb has been successfully operational since late 2003, responding to five different science disciplines and acquiring data from over 10 different sources. Table1 displays a list of the science tracking system integrated into our system. ... ..."

"... In PAGE 18: ... The simplest might be to give the student the correct answer at this point and go on to the next item on the agenda. Learning Outcomes Analysis of the pre- and post-test results clearly demonstrate that use of CIRCSIM-Tutor for 1 hr results in learning about the baroreceptor reflex and a greater ability to use this understanding to carry out qualitative reasoning about the system (see Table1 ). Students were able to correctly describe more of the relationships between system variables (Part 1), with scores increasing from 13.... ..."

"... In PAGE 111: ... 5.11 What Type of Hint to Provide? Table7 shows the frequency of initial hints (PT-Hint versus CI-Hint) provided in a string of hints by JAM and AAR. JAM has a slight tendency to initially use a CI-Hint while AAR has the opposite tendency.... ..."

"... In PAGE 5: ... The log files from these sessions contained 9094 student utterances, each of which was classified by an expert. The expert ratings were com- pared to the classifier apos;s ratings, forming a 2 x 2 contin- gency table for each category as in Table4 . To expedite ratings, utterances extracted from the log files were split into two groups, contributions and non-contributions, according to their logged classifica- tion.... In PAGE 6: ... High inter-rater reli- ability on the monothetic classification task renders polythetic schemes superfluous for this application. The recall column for evaluation in Table4 is gener- ally much higher than corresponding cells in the preci- sion column. The disparity implies a high rate of false positives for each of the categories.... ..."

"... In PAGE 9: ... The dependent measures were different for computer lit- eracy and physics, so the two sets of studies will be dis- cussed separately. Table1 presents data on three experiments on computer literacy. The data in Table 1 constitute a reanalysis of studies reported in two published conference proceed- ings (Graesser, Moreno, et al.... In PAGE 9: ... Table1 presents data on three experiments on computer literacy. The data in Table1 constitute a reanalysis of studies reported in two published conference proceed- ings (Graesser, Moreno, et al., 2003; Person, Graesser, Bautista, et al.... In PAGE 10: ...loze tasks. The Graesser, Moreno, et al. study had the same design except that a pretest was administered, there was an expanded set of deep multiple-choice questions, and a textbook-reduced comparison condition was used instead of the textbook condition. In Table1 , the means, standard deviations, and effect sizes in standard deviation units are reported. The effect sizes revealed that AutoTutor did not facilitate learning on the shallow multiple-choice test questions that had been prepared by the writers of the test bank for the text- book.... In PAGE 10: ... All of these effect sizes were low or negative (mean effect size was .05 for the seven values in Table1 ). Shal- low knowledge is not the sort of knowledge that AutoTutor was designed to deal with, so this result is not surprising.... In PAGE 10: ...f .50. These effect sizes were generally larger when the comparison condition was read nothing (M H11005 .43 for nine effect sizes in Table1 ) than when the comparison condition was the textbook or textbook-reduced condition (M H11005 .22).... ..."