The proactive retrieval system is interrupting my writing

Recall recognition Writing title No interruption 108.10(27.79) 118.88(48.78) 22.50(8.19) 24.00(8.10) Related Interruption 99.00(38.19) 93.33(61.97) 24.25(14.56) 20.00(5.09) Not related interruption 101.25(45.06) 103.50 (60.10) 44.40(45.52) 24.00(9.84) Interruption task. We also analyzed if there were differences on time performance of the interruption task (TOI) in the related and not related conditions (averages of 77.00 and 64.50 seconds respectively). The differences were not significant F(1,30)=1.62, p=0.21. Perceived Quality of the Performance of the Main Task. Analysis of the questions showed significant differences as a function of the Main task F(3,66)=5.10, p<0.005 and of the interruption task F(2,66)=5.69, p<0.005. Table2. Average and standard deviation (between parentheses) of perceived quality of performance of the main tasks according to the interruption tasks conditions abstract recall recognition Writing titlerecall recognition Writing title No interruption 3.00(0.67) 3.40(0.96) 3.6(1.18) 4.50(0.76) Related Interruption 4.2(0.96) 3.5(0.58) 4.2(0.83) 4.80(0.44) Not related interruption 2.75(0.95) 3.5(1.00) 3.00(1.00) 3.60(0.91) For the main tasks, pairwise comparisons showed that writing the title was significantly different than the rest (all p<0.005). As expected, participants were more satisfied with their performance in this easy task than with the others. No other comparisons were significant. The differences between the conditions without interruption and the conditions with related interruptions was significant. Differences between related and not related interruptions were also significant (both p<0.005). These results indicate that there is not perceived difference between the conditions without interruption and with not related interruptions. In the related condition participants considered the quality of the main task was higher. 4 DISCUSSION AND CONCLUSIONS Proactive Retrieval systems (PRs) such as IntelliGent retrieve and decide which available information is most likely relevant to the text being written, and present it without user intervention. In this paper we present the results of a preliminary study in which the effects of presenting peripheral information are explored. The lack of significant differences in the main task performance between interrupted and not interrupted tasks suggests that resuming the writing task after an interruption from IntelliGent does not seriously impair time task performance. Furthermore, results from the questionnaire show that participants judged the quality of their performance better when a related paper was presented than in the conditions in which the paper was unrelated or IntelliGent was not present. We could conclude then that IntelliGent does not require extra time to resume the interrupted writing task and, when relevant information is presented, users are more satisfied with the quality of their performance. This results seem to conflict with our previous findings [3] in natural working environments. Users felt that the disruptions of the system affected their writing performance and these disruptions were negatively perceived as very frequently the system offered paper not relevant to the contents of the written text. This discrepancy in results can be explained in two related ways. First, in our first study, users were university researchers and were asked to use the system while writing their scientific papers. This writing task seems to have more cognitive load than any of the tasks participants were confronted in this experiment and, consequently can be more sensitive to interruptions than the tasks we used in the present study. Second, this study was designed to get an impression of how IntelliGent could affect several writing tasks. It would be interesting to analyze the interactions between the complexity of the tasks and types of interruptions in detail. However, doing this in a within-subjects design would require that all subjects perform such a large number of tasks that it becomes questionable whether they can complete the experiment without significantly changing their attitude and behaviour during the course of the experiment. In addition, it might be difficult to construct tasks with a sufficiently similar degree of difficulty for all conditions. Alternatively, one might opt for a between-subjects design, accepting the risk that medium size effects may fail to reach significance because of large differences between the cognitive capabilities of the subjects. Perhaps the best solution is to reduce the number of levels in the ‘task type’ factor to two. Since we found that writing tasks with higher cognitive load such as writing abstracts or answering questions, require more time to be performed than simpler tasks, we plan to fully factorial within subjects experiments. In one we will compare abstract and recall and in another we will compare recall and recognition. The first experiment would gauge the effect in two lengthy tasks, whereas the second would give insight into the differences between a long and complex tasks and a short and simple one. ACKNOWLEDGMENTWe would like to thank the Search Expertise Centrum, Utrecht, The Netherlands for providing us withthe ‘IntelliGent’ system and IOP SenterNovem, The Hague, The Netherlands, for funding this project. REFERENCES[1] Bailey, B.P., Konstan, J, and Carlis, J.V. (2000). Measuring the effects of interruptions on task performance in theuser interface. In: IEEE Conferences on Systems, Man and Cybernetics, 752-762.[2] Budzik, J. and Hammond, K. (1999). Watson: Anticipating and Contextualizing Information Needs. Proc. 62ndAnn. Meeting Am. Soc. for Information Science, 727-740.[3] Deshpande, A., Boves, L. and Puerta Melguizo, M.C. (2006). À propos: Pro-active personalization for professionaldocument writing. SigWriting: 10 International Conference of the EARLU special interest group on writing.[4] Hayes, L.S. and Flower, J.R. (1980). Identifying the organization of writing processes. IN: Gregg, L.W. &Steinberg, E.R. eds. Cognitive Processes in Writing, Hillsdale, NJ, Lawrence Erlbaum, 3-30.[5] Lieberman, H. (1997) Autonomous Interface Agents. ACM Conference on Human-Computer Interface, Atlanta, 67-74.[6] Maglio, P.P. and Campbell, C.S. (2000). Tradeoffs in displaying peripheral information. Proceedings of the CHI2000 conference on Human factors in computing systems, New York: ACM Press, 241-248.[7] Piolat, A., Kellogg, R.T., and Farioli, F. (2001) The triple task technique for studying writing processes: On whichtask is attention focused? Current Psychology Letters: Behaviour, Brain & Cognition, 4, 67-83.[8] Rhodes, B.J. (2000). Just-in-time Information Retrieval, Phd Thesis, MIT.[9] Search Expertise Centrum, Consultancy B.V. Utrecht, The Netherlands.[10] Shimamura, A.P. and Squire, L.R. (1998). Long-term memory in amnesia: Cued recall, recognition memory andconfidence ratings. Journal of Experimental Psychology: Learning, memory and cognition, 14, 763-770.[11] Zijlstra, F.R.H., Roe, R.A., Leonora, A.B., and Krediet, I. (1999). Temporal Factors in Mental Work: Effects ofInterrupted Activities. Journal of Occupational and Organizational Psychology, 72, 163-185.