THE "STRUCTURAL STANDARDS" APPROACH TO HUMAN ERROR

This paper suggests a theoretical model of "structural standard" which depicts human activity in its socio-technical context. Based on the model, it re-defines errors as performance which deviates from technical and/or social norms of acting and deduces the strategy of error management for evolutionary iterative design and prototyping,

INTRODUCTION

Most researchers and practitioners agree that error can be defined as a performance deviating from  the act which should have been performed (Senders and Moray, 1991). This definition implies: (1) an act’s standard, or norm (Meister and Rabideau, 1965; Groeger, and Chapman 1990); (2) actual performance of the act which deviates from the standard; (3) identification of the performance as a specific error type by an analyst judging the performance against the standard (Leplat, 1990); and (4) standards of error analysis itself (Rasmussen, 1990). Considering these implications, this paper suggests a theoretical model of "structural standard" which depicts an individual act in its socio-technical context and interprets errors as deviations from technical and/or social norms of acting. Then it deduces the strategy for error management from the model and suggests a typology for the error management devices.

THE “STRUCTURAL STANDARD” MODEL

Standard types of performance and re-definition of errors

The structural standard model is based on the concepts of norm and deviation used in the social sciences for many years. Norm can be defined as “... any standard or rule that states what human beings should or should not think, say, or do under given circumstances" (Blake and Davis, 1964, p.456). I suggest that “standard” is a more basic term than “norm”. According to the structural standard model, norms and deviations of an act are two out of four different types of the act’s standards. To represent a deviation as a standard one must: (1) identify and label a performance as a certain type of deviation from the act’s norm and (2) identify normal technical results of and/or norms of social responses to this type of deviant performance of the act.. Structural standard of an act consists of the following four standard types associated with norms for technical results and social responses (Figure 1).

Ideal-norm act and correct performance. There are acts which elicit "maximum potential return", corresponding to "ideal performance" (Jackson, 1966). It can be the achievement of some fixed value (e.g., 0-defect level), or performing the best, relative to other actors (e.g., winning the Olympic Gold Medal), or performing at the "golden middle" level (e.g., appropriate level of participation in group discussion earning the highest approval by the group). In all cases, the ideal or close to ideal performance ought to produce optimal technical results and/or to elicit social reward. When errors are concerned, the ideal-norm act is the one which is usually considered correct.

Figure 1. The structural standard model.

Regular-norm act and potential errors. Regular-norm can be viewed as an ideal norm loosened by tolerance. Tolerance defines limits of acceptable deviation from the ideal norm and is a norm itself. A regular-norm performance ought to produce acceptable technical results and/or to elicit a regular positive social response. This regular response should be distinguished from award associated with an ideal act. For example, a salesperson receives regular commissions for a regular norm act of  sales satisfying the quota, but is awarded a bonus for an ideal-norm act of substantially exceeding the quota. Other important components of the technical or social response to a regular-norm act are tolerance absorbing and tolerance offsetting. For example, in manufacturing, worker time allowances (idle time and delays) are offset by assigning more workers for the same job. Regular-norm performance usually is considered as a correct one because it produces acceptable results. At the same time, performance that is close to the tolerance limits under the changed circumstances can result in a failure (Rasmussen, 1986), thus constituting a potential error. Sometimes tolerant deviations accumulate or combine with other factors and lead to technical failure or social deviance. In this case a sequence of potential errors is itself a factor of the situation change. J. Reason (1990) calls such regular-norm acts latent errors .

Remedied-violation act and recovered errors. In many cases, performance of an act which deviates from the norm beyond the tolerance limits can still be corrected. Remedied violation is produced by a remediable deviation followed by a remedial act performed by the actor. Remedial act ought to  prevent technical failure and/or to countermand a negative social sanction. There are norms for remedial acts appropriate for each situation of remediable violation. For example, one English For Foreigners course explains: "If you are ten minutes late for a meeting, you should say 'I am sorry I am late'. If you are fifteen minutes late, you should say 'I am very sorry I am late'." Remedied-violation performance usually is not considered as an error because it does not have negative consequences. Kantowits and Sorkin (1983) call such performance recovered errors. They suggest that recovered errors should be addressed by designers “because they do reveal design inadequacies. Today's recovered error may be tomorrow's disaster." (Kantowitz and Sorkin, 1983; p. 32).

Violation act and errors proper. A violation is either a remediless deviation beyond the tolerance limits or a remediable deviation not followed by an appropriate remedial act. The result ought to be technical failure and/or negative social sanction. What is usually called an error can be characterized as a violation act.

Interpreting given performance, its technical results and social responses to it in terms of the structural standard model, an analyst can determine a normative status of the performance (ideal-norm, regular-norm, remedied violation, or violation) and then address it appropriately. Normative status of an act depends not only on the performance itself but also on the socio-technical context. For example, same performance can be a regular-norm one (a correct act) in “tolerant-supportive” system and a violation (an error) in “restrictive-punitive” system (Jackson, 1966).

Strategies and means for improving  human performance

The model suggests a general strategy for improving human performance: upgrading normative status of performance by stretching system’s tolerance and remedy limits, on one hand, and supporting better performance on the other. The model also suggests the appropriate technical means. Below are examples of means which are appropriate for upgrading simple operations.

Upgrading violations to remedied violations. The definitions of violation and remedied violation suggest main approaches to upgrading violations to remedied violations: (1) preventing a violation by warning of consequences, making it difficult or even impossible, e.g., by “forcing functions” (Norman, 1990); (2) widening remedial limits, e.g., by providing additional reversals, such as multiple-step undoes; (3) preventing a failure to perform remedial act, e.g., by highlighting negative results, informing an actor of possible remedies, prompting and/or suggesting remedial act.,.

Upgrading remedied violations to regular?norm acts. The definitions of remedied violation and regular-norm acts suggest that upgrading a remedied violation to a regular-norm act can be achieved by automating the remedial act. The automation transforms a remedial act formerly performed by an actor into a system process offsetting the deviation. An example is a computer system automatically repairing and then performing a misspelled command, instead of just informing a user about the error and prompting to re-enter a command correctly.

Upgrading regular-norm acts to ideal-norm acts. A system designer should always look for any possibility of making specific tolerances unnecessary, thus upgrading a regular-norm act to the ideal-norm one. A classic example is using command menus to avoid misspelling of the commands.

The above strategies of continuous performance upgrading are especially well suited for evolutionary iterative design or prototyping. Changes in design of a system may affect  tolerance and remedy limits.  Therefore, special care should be made to prevent  degradation of normative status of performance. The means are essentially the same. The strategy is reviewing every change of the design for possible negative side-effects on normative status of main operations and, if any, providing for upgrading them at least to the previous level.

CONCLUSION

This paper has re-defined human errors as deviations from technical and social norms of acting according to the structural standard model of human act. The model distinguishes four normative statuses of act’s performance: an ideal (correct act), regular act (potential error), remedied violation (recovered error), and violation (error proper). The paper has illustrated how the strategies for error management can be deduced from the model and can be used in the process of iterative design and prototyping. The model also suggests the typology for technical means of error handling. The concepts of tolerance and remedy limits, tolerance absorbing and offsetting, and remedial acts are especially relevant to the modern workplace with its "fuzzy" task definitions. Actors should be allowed to select performance strategies via trial-and-error exploration of the system's functionality and limits (Rasmussen, 1990). Modern systems should be tolerant-supportive with extended tolerance limits and corresponding extended tolerance absorbing capacities and offsetting devices; with extended remedy limits and corresponding variety of possible remedial acts.

REFERENCES

BLAKE, J. and DAVIS, K. (1964). Norms values and sanctions. In R.E. Faris (Ed.). Handbook of Modern Sociology. (Rand McNally, Chicago), 456-484.

GROEGER,J.A. and CHAPMAN,P.R. (1990). Errors and Bias in assessments of danger and frequency of traffic situations. Ergonomics, vol. 33, no. 10-11,1349-1363.

JACKSON, J. (1966). A conceptual and measurement model for norms and roles. Pacific Sociological Review, no. 9, 35?47.

KANTOWITZ, B.H. and SORKIN, R.D. (1983). Human Factors: Understanding People?System Relationships. (Wiley & Sons, New York)

LEPLAT, J. (1990). Relations between task and activity: elements for elaborating a framework for error analysis. Ergonomics, vol. 33, no. 10-11, 1389-1401.

MEISTER, D. and RABIDEAU, G.R. (1965). Human Factors Evaluation in System Development. (Wiley,New York)

NORMAN, D.A. (1990). Design of Everyday Things.(Doubleday/Currency, New York).

RASMUSSEN, J. (1986). Information Processing and Human?machine Interaction: An Approach to Cognitive engineering. North Holland, Amsterdam.

RASMUSSEN, J. (1990). The role of error in organizing behavior. Ergonomics, vol. 33, no. 10-11, 1185-1199.

REASON, J. (1990). Human Error. (Cambridge University Press, Cambridge).

SENDERS, J.W. and MORAY, N.P. (1991). Human Error. (Lawrence Erlbaum Associates, Hillsdale.