Abstract
A generally intelligent machine (AGI) should be able to learn a wide range of tasks. Knowledge acquisition in complex and dynamic task-environments cannot happen all-at-once, and AGI-aspiring systems must thus be capable of cumulative learning: efficiently making use of existing knowledge during learning, supporting increases in the scope of ability and knowledge, incrementally and predictably — without catastrophic forgetting or mangling of existing knowledge. Where relevant expertise is at hand the learning process can be aided by curriculum-based teaching, where a teacher divides a high-level task up into smaller and simpler pieces and presents them in an order that facilitates learning. Creating such a curriculum can benefit from expert knowledge of (a) the task domain, (b) the learning system itself, and (c) general teaching principles. Curriculum design for AI systems has so far been rather ad-hoc and limited to systems incapable of cumulative learning. We present a task analysis methodology that utilizes expert knowledge and is intended to inform the construction of teaching curricula for cumulative learners. Inspired in part by methods from knowledge engineering and functional requirements analysis, our strategy decomposes high-level tasks in three ways based on involved actions, features and functionality. We show how this methodology can be used for a (simplified) arrival control task from the air traffic control domain, where extensive expert knowledge is available and teaching cumulative learners is required to facilitate the safe and trustworthy automation of complex workflows.
The authors gratefully acknowledge partial funding for this project from Isavia, IIIM, Reykjavik University, Delft Univeresity of Technology, and the Netherlands Organization for Scientific Research (NWO grant 313-99-3160/Values4Water project).
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsNotes
- 1.
Negative transfer of training may also occur, where pre-existing knowledge interferes with learning something new — e.g. a racquetball player may take longer to get used to the way a squash ball bounces than somebody who never played racquetball. An optimal curriculum would mitigate negative transfer as much as possible.
- 2.
However this is measured, we expect at a minimum the ‘learning cycle’ (alternating learning and non-learning periods) to be free from designer intervention at runtime. Given that, the smaller those periods become (relative to the shortest perception-action cycle, for instance), to the point of being considered virtually or completely continuous, the better the “learning always on” requirement is being met.
- 3.
For instance, they typically require up-front full data disclosure (final data set up-front), all-at-once training (to train on the final data set from the very beginning) and learning-free deployment (the need to turn off learning before deployment to avoid unpredictable drift; cf. [15]).
- 4.
The ADDIE model for instructional design consists of (1) analysis of the learner, learning goals, and teaching constraints, (2) design of the lesson plan or curriculum, which involves subject matter/task analysis, (3) development or assembly of the actual training materials, (4) implementation of the instruction with the learner (i.e. the actual teaching/training/learning), and (5) evaluation of learning outcomes.
- 5.
Due to space limitations we only describe a highly simplified version of arrival control here. A more elaborate version can be found in our tech report: http://www.ru.is/faculty/thorisson/RUTR18001_ArrivalControl.pdf.
- 6.
Isavia is Iceland’s aviation authority, managing air traffic in an area measuring 5.4 million square kilometers.
- 7.
A lot has even been written on task analysis for ATC (cf. https://www.eurocontrol.int/articles/atco-task-analysis), but we still need a new method for designing curricula for non-human cumulative learners.
- 8.
A more elaborate version can be found in our tech report: http://www.ru.is/faculty/thorisson/RUTR18001_ArrivalControl.pdf.
References
Amodei, D., Olah, C., Steinhardt, J., Christiano, P.F., Schulman, J., Mané, D.: Concrete problems in AI safety. CoRR abs/1606.06565 (2016)
Barry, A.: A hierarchical XCS for long path environments. In: Proceedings of GECCO-2001 (2001)
Bengio, Y., Louradour, J., Collobert, R., Weston, J.: Curriculum learning. In: Proceedings of ICML-26 (2009)
Bieger, J.: Artificial pedagogy: a proposal. In: HLAI 2016 Doctoral Consortium, New York, NY (2016)
Bieger, J., Thórisson, K.R., Garrett, D.: Raising AI: tutoring matters. In: Goertzel, B., Orseau, L., Snaider, J. (eds.) AGI 2014. LNCS (LNAI), vol. 8598, pp. 1–10. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-09274-4_1
Bieger, J., Thórisson, K.R., Steunebrink, B.R.: The pedagogical pentagon: a conceptual framework for artificial pedagogy. In: Everitt, T., Goertzel, B., Potapov, A. (eds.) AGI 2017. LNCS (LNAI), vol. 10414, pp. 212–222. Springer, Cham (2017). https://doi.org/10.1007/978-3-319-63703-7_20
Branson, R.K., Rayner, G.T., Cox, J.L., Furman, J.P., King, F.J.: Interservice Procedures for Instructional Systems Development: Phase 4 and 5. Florida State University, Technical report (1975)
Burke, L.A., Hutchins, H.M.: Training transfer: an integrative literature review. Hum. Resour. Dev. Rev. 6(3), 263–296 (2007)
Hasselmo, M.E.: Avoiding catastrophic forgetting. Trends Cognit. Sci. 21(6), 407–408 (2017)
Hengst, B.: Hierarchical approaches. In: Wiering, M., van Otterlo, M. (eds.) Reinforcement Learning. Adaptation, Learning, and Optimization, vol 12. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-27645-3_9
Kirkpatrick, J., Pascanu, R., Rabinowitz, N., Veness, J., Desjardins, G., Rusu, A.A., Milan, K., Quan, J., Ramalho, T., Grabska-Barwinska, A.: Overcoming catastrophic forgetting in neural networks. PNAS 114(13), 3521–3526 (2017)
Lake, B., Salakhutdinov, R., Gross, J., Tenenbaum, J.: One shot learning of simple visual concepts. In: Proceedings of CogSci 2011, vol. 33 (2011)
Looks, M.: Competent program evolution. Ph.D. thesis, Washington University (2006)
Lu, J., Behbood, V., Hao, P., Zuo, H., Xue, S., Zhang, G.: Transfer learning using computational intelligence: a survey. Knowl. Based Syst. 80, 14–23 (2015)
Marcus, G.: Deep learning: a critical appraisal. CoRR abs/1801.00631 (2018)
Nivel, E., Thórisson, K.R., Steunebrink, B.R., Dindo, H., Pezzulo, G., Rodriguez, M., Hernandez, C., Ognibene, D., Schmidhuber, J., Sanz, R., Helgason, H.P., Chella, A., Jonsson, G.K.: Bounded recursive self-improvement. Technical RUTR-SCS13006, Reykjavik University, Reykjavik, Iceland (2013)
Pollock, J.L.: Defeasible reasoning and degrees of justification. Argum. Comput. 1(1), 7–22 (2010)
Schreiber, G.: Knowledge Engineering and Management: The Common KADS Methodology. MIT Press, Cambridge (2000)
Silver, D.L., Yang, Q., Li, L.: Lifelong machine learning systems: beyond learning algorithms. In: AAAI Spring Symposium: Lifelong Machine Learning (2013)
Skinner, B.F.: The Behavior of Organisms: An Experimental Analysis. Appleton-Century-Crofts Inc., New York (1938)
Steunebrink, B.R., Koutník, J., Thórisson, K.R., Nivel, E., Schmidhuber, J.: Resource-bounded machines are motivated to be effective, efficient, and curious. In: Kühnberger, K.-U., Rudolph, S., Wang, P. (eds.) AGI 2013. LNCS (LNAI), vol. 7999, pp. 119–129. Springer, Heidelberg (2013). https://doi.org/10.1007/978-3-642-39521-5_13
Steunebrink, B.R., Thórisson, K.R., Schmidhuber, J.: Growing recursive self-improvers. In: Steunebrink, B., Wang, P., Goertzel, B. (eds.) AGI -2016. LNCS (LNAI), vol. 9782, pp. 129–139. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-41649-6_13
Teague, R.C., Gittelman, S.S., Park, O.c.: A review of the literature on part-task and whole-task training and context dependency. Technical report, ARI, US (1994)
Teh, Y.W., Bapst, V., Czarnecki, W.M., Quan, J., Kirkpatrick, J., Hadsell, R., Heess, N., Pascanu, R.: Distral: robust multitask reinforcement learning. CoRR abs/1707.04175 (2017)
Thórisson, K.R., Bieger, J., Thorarensen, T., Sigurðardóttir, J.S., Steunebrink, B.R.: Why artificial intelligence needs a task theory. In: Steunebrink, B., Wang, P., Goertzel, B. (eds.) AGI -2016. LNCS (LNAI), vol. 9782, pp. 118–128. Springer, Cham (2016). https://doi.org/10.1007/978-3-319-41649-6_12
Vygotsky, L.S.: Interaction between learning and development. In: Cole, M., John-Steiner, V., Scribner, S., Souberman, E. (eds.) Mind in Society: The Development of Higher Psychological Processes. Harvard University Press, Cambridge (1978)
Wood, D., Bruner, J.S., Ross, G.: The role of tutoring in problem solving. J. Child Psychol. Psychiatry 17(2), 89–100 (1976)
Zhan, Y., Taylor, M.E.: Online transfer learning in reinforcement learning domains. CoRR abs/1507.00436 (2015)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2018 Springer Nature Switzerland AG
About this paper
Cite this paper
Bieger, J.E., Thórisson, K.R. (2018). Task Analysis for Teaching Cumulative Learners. In: Iklé, M., Franz, A., Rzepka, R., Goertzel, B. (eds) Artificial General Intelligence. AGI 2018. Lecture Notes in Computer Science(), vol 10999. Springer, Cham. https://doi.org/10.1007/978-3-319-97676-1_3
Download citation
DOI: https://doi.org/10.1007/978-3-319-97676-1_3
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-97675-4
Online ISBN: 978-3-319-97676-1
eBook Packages: Computer ScienceComputer Science (R0)