Studies of human error
Human Error Modelling (HUM) Project
Studies of human error
Actions are frequently required that do not move an individual towards a goal state but remain critical to performing a task correctly. We refer to these actions as being device-specific. Our work has shown that even when an individual is actively engaged in avoiding these types of errors they still systematically fail . The empirical work being conducted as part of the HUM project aims to discover why device-specific errors are so difficult to avoid. The extent to which individuals were able to self-report device-specific errors was investigated . It was found that the pervasiveness of these types of errors was recognisable but underlying cognitive and attentional causes were not. Forgetting to perform an action can only be reported before a cue (used to trigger future actions) has decayed from memory. A cue can be visual (e.g., the strategic positioning of a post-it note on a desktop) or associative (e.g., remembering to clean shoes may be triggered by an intention to leave for work). This suggests that consideration of how cues are used within an environment should be considered alongside cognitive explanations.
Some of our work has focused on understanding post-completion errors (PCEs), a class of errors in which a terminating step is required – but omitted – after the main goal of an interaction is achieved. Previous laboratory studies on PCE have found during routine procedural tasks the frequency of the error can be lowered by providing just-in-time cues  or reducing working memory demands . Our work has built on these findings and shown that when the process of retrieving previously formulated from memory is interrupted  or is more difficult due to increased task complexity , PCEs are significantly more likely. We have also demonstrated that PCEs can also occur in non-routine problem-solving situations  and that using certain types of just-in-time cues are more effective than others .
Recent experimentation compared two error classes: omission and mode errors . It was found that omission errors are trigged by the visual and cognitive salience of a subsequent procedural step that captures attention. When task complexity was high omission errors were significantly more likely. Mode errors occur when an individual fails to perform an appropriate action sequence. It was found that mode errors were only significantly more likely when trial complexity was high and the visual salience of a critical device signal was low (due to an increased level of perceptual load). This suggests that studying error classes, where multiple action sequences are possible, is likely to require consideration of environmental load variables when developing a causal account.
We believe that the pervasiveness of device-specific errors is due to a preference to rely on cues within the environment to drive an interaction. The tendency to rely on environmental cues instead of internalised associative cues might be explained by a theory of attentional activation. Our aim is to develop an empirically derived model of interaction that explains how the cognitive salience and semantic properties of environmental cues influence the systemacity of errors. We have developed novel ways of manipulating environmental conditions so that causal explanations of human error can be developed . While it is valid to suggest that errors are more likely when the load placed on cognitive resources is high or when an individual is interrupted, depending on the way attention is controlled error manifestations can arise from: stimulus-driven confusions, stimulus-driven misinterpretations, goal-driven confusions, or goal-driven misinterpretations . This approach enables questions such as ‘Where do I start interacting?’ and ‘What should I be attending to?’ to be addressed enabling a deeper understanding of human error.
1. Back, J., Cheng, W.L., Dann, R., Curzon, P., & Blandford, A. , Does being motivated to avoid procedural errors influence their systematicity? To appear in Proceedings of HCI 2006.
2. Back, J., Blandford, A. & Curzon, P. Self-reporting Erroneous and Exploratory Interactions. Submitted.
3. Chung, P. & Byrne, M. D. , Visual cues to reduce errors in a routine procedural task. Proceedings of the 26th Annual Conference of the Cognitive Science Society. Hillsdale, NJ: Lawrence Erlbaum Associates.
4. Byrne, M. & Bovair, S. , A Working Memory Model of a Common Procedural Error. Cognitive Science 21(1), 31-69.
5. Li, S., Cox, A., Blandford, A., Cairns, P., & Abeles, A. , Further investigations into post-completion error: the effects of interruption position and duration. To appear in Proceedings of the 28th Annual Meeting of the Cognitive Science Society, Vancouver, BC, Canada, July 26-29, 2006.
6. Li, S., Blandford, A., Cairns, P. & Young, R. M. , Post-completion errors in problem solving. Proceedings of the 27th Annual Conference of the Cognitive Science Society. Hillsdale, NJ: Lawrence Erlbaum Associates.
7. Back, J., Blandford, A. & Curzon, P. Exploring The Systemacity Of Procedural Errors. In preparation.
8. Back, J., Blandford, A. & Curzon, P. A load Theory for Predicting Errors. In preparation.
9. Back, J., Blandford, A. & Curzon, P. A Cognitive Load Theory for Human Error: The Modelling Of Activation. In preparation.
This page last modified
11 September, 2007
by George Papatzanis