Sunday, April 6, 2014

NRC Issues Safety Culture Common Language NUREG

The NRC has issued NUREG-2165* which formalizes the safety culture (SC) common language that has been under development since the NRC SC Policy Statement (SCPS) was issued.  On topics important to us the NUREG repeats word-for-word the text of a document** prepared after a common language workshop held January 29-30, 2013.  Both documents contain a set of SC traits, attributes that define each trait and examples that would evidence each attribute.  Because the language is the same, our opinion on the treatment of our important topics remains the same, as described in detail in our Feb. 28, 2013 post.  Specifically, the treatment of

Decision making, including the treatment of goal conflicts, is Good;

Corrective action, part of problem identification and resolution, is Satisfactory;

Management Incentives is Unsatisfactory because the associated attributes focuses on workers, not managers, and any senior management incentive program is not mentioned; and

Work Backlogs are mentioned in a couple of specific areas so the overall grade is Minimally Acceptable.


But we have one overarching concern that transcends our opinion of common language specifics.


Our Perspective

Our biggest issue with the traits, attributes and examples approach is our fear it will lead to the complete bureaucratization of SC evaluation, either consciously or unconsciously.  The examples in particular can morph into soft requirements on a physical or mental checklist.  Such an approach leads to numerous questions.  How many of the 10 traits does a healthy or positive SC exhibit?***  How many of the 40 attributes?  Are the traits equally important?  How about the attributes?  Could the weighting factors vary across plant sites?  How many examples must be observed before an attribute is judged acceptably present?

We understand the value of effective communications among regulators, licensee personnel and other stakeholders.  But we worry about possible unintended consequences as people attempt to apply the guidance in NUREG-2165, especially in the NRC’s Reactor Oversight Process (ROP).****


*  NRC NUREG-2165, “Safety Culture Common Language” (Mar. 2014).  ADAMS ML14083A200.

**  Nuclear Safety Culture Common Language 4th Public Workshop January 29-31, 2013.  ADAMS ML13031A343.

***  The NUREG-2165 text describes a “healthy” SC while the SCPS (published as NUREG/BR-0500, Rev. 1, ADAMS ML12355A122) refers to a “positive” SC.  The correct answer to “how many traits?” may be “more than ten” because the authors note “There may also be traits not included in the SCPS that are important in a healthy safety culture.” (p. 2)

****  The common language “initiative is within the Commission-directed framework for enhancing the ROP treatment of cross-cutting areas to more fully address safety culture.” (p. 3)  This may require a little linguistic jujitsu since the SCPS says “traits were not developed for inspection purposes.”

1 comment:

  1. Lew,

    As an old mutual friend used to say from time to time "It's never too late to give up hope."

    The history of behaviorist conditioning experiments over the entire 20th Century amply confirms that itemization begets standardized metric approaches to performance assessment - which are inevitably extended beyond the bounds of their applicability.

    What may not be obvious is that the safety culture assessment tools that have been in use for two decades now are what conditioned this latest action. Much as no one questions if the marks are properly placed on a machine manufactured yard stick, neither do "safety culture experts" in Nuclear question the validity of NSC surveys. But they should.

    One can see from its initial structure, the attempt with the 1999 ROP to keep management judgment issues (i.e. via the Cross-cutting areas) with a non-metric place in the overall assessment - even as the quantified "cornerstones" came to take the place of the more judgmental desicions processes in the SALP. This partition is likely to evaporate.

    The inherent fragmentation of performance (cf. Sonia Haber's landmark NUREG/CR- 5538, circa 1991), absent the strong influence of reintegrating principle sets (e.g. of the sort espoused by Rickover in 1983, and others since (such as Covey)) means that "counting stuff" becomes the default performance measure.

    There are two problems commingled here: 1) Engineers and Operators are naturally drawn to the kinds of psychological comfort counting things brings. When surrounded by artifacts with lots of things that can be reliably counted and trended, they might not realize that some important measures of performance are non-metric (e.g. distraction from anxiety about an ill family member); 2) Many clinical psychologist researchers suffer from engineering-envy and a prone to design experiments and draw conclusions suitable for stimulating that same kind of "counting comfort."

    I'd say that this represents the near final "victory" in the long battle by the industry to shift the burden of disaster level risk onto the NRC and the public - hence forth, no quantity-justified citation means no finding of defect. Sad.

    ReplyDelete

Thanks for your comment. We read them all. We'd like to display them under their respective posts on our main page but that's not how Blogger works.