Thought Bias: The Hidden Pipeline Integrity Threat Michael Rosenfeld and Joel Anderson RSI Pipeline Solutions LLC Pipeline Pigging and Integrity Management Conference February 6-10, 2023 Organized by Clarion Technical Conferences Pipeline Pigging and Integrity Management Conference, Houston, February 2023 Proceedings of the 2023 Pipeline Pigging and Integrity Management Conference. Copyright ©2023 by Clarion Technical Conferences and the author(s). All rights reserved. This document may not be reproduced in any form without permission from the copyright owners. 2 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 Abstract Pipeline regulations, industry standards, and technical research set forth extensive guidance for managing threats to pipeline integrity through a formal integrity management (IM) plan. Such plans rely on a rigorous procedural approach to identify threats and mitigate risk on a prioritized basis in a systematic and repeatable process. One subtle threat that is not just overlooked but is almost invisible to many integrity management personnel is that of biased thought processes. Because they typically go unrecognized such biases can seriously undermine the effectiveness of IM programs in a variety of ways that lead to poor decisions. Such biases may also affect routine pipeline construction and maintenance projects outside of IM work, but which may lead to harmful long-term IM implications. Even when information to the contrary exists prior to the decision, people can become anchored to a fallacy, unwilling to move from it. The various forms of bias, examples of their potential adverse effects on pipeline integrity, warning signs, and potential avoidance methods are discussed. Introduction Pipeline integrity management standards and regulations list the integrity threats to be managed, mitigated, or prevented by integrity management plans. Almost all recognized threats are metallurgical or mechanical in nature, for example corrosion, seam fatigue, earth movement, or equipment failure. They can be managed with specifications, inspections, tests, or other engineering barriers against failure. The exception is “operator error”, which occurs when an operator responds incorrectly when presented with information about a potential problem and the resulting decision, action, or inaction is directly responsible for the failure. Operator error can be managed with administrative barriers such as procedures or worker training.1 The authors have observed that many pipeline incidents are more complex than what is implied by the simple direct cause. Complexity arises in the form of interactions of physical factors in about 7% of reportable incidents.2 “Interaction” is a combination of factors that result in a more severe condition or higher probability of failure than the individual factors considered separately. Operators often fail to address interactions because they overlook the possibility of their occurrence and therefore fail to collect the necessary data to reveal the potential for interaction. One or more types of thought bias may lead to such a sequence. Sometimes operators experience multiple incidents or near-miss events that are seemingly unrelated, but such patterns are indicative of an underlying thought bias that prevents the operator from connecting the events. It is the authors’ opinion that thought bias can negate the effectiveness of a formal IM plan no matter how detailed or well thought out, lead to unsafe practices in routine work, or counteract best 1 One could take the reasonable position that all failures are due to underlying operator error. Consider a failure due to external corrosion, a naturally occurring physical process. The direct cause is a double failure of the coating system and the cathodic protection (CP) system to perform adequately, but the underlying causes likely originated with people. Was the coating applied properly? Was the pipeline carefully lowered into the ditch and inspected diligently? Was the CP system adequately specified and its performance monitored? Was the in-line inspection properly interpreted and acted on? An answer to “why” for any of those factors is the essence of a root cause. They originated with people; physics is not at fault. 2 Munoz, E. and Rosenfeld, M.J., “Improving Models to Consider Complex Loadings, Operational Considerations, and Interactive Threats”, Task III.B.3 Final Report, US DOT Contract DTPH56-14-H00004, Dec. 30, 2016. 3 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 intentions in pipeline projects. Moreover, such bias appears to be pervasive within the industry. What thought bias is, how it can manifest itself to adversely affect pipeline integrity management, and strategies to discover and mitigate it are discussed in this paper. What is Thought Bias? Heuristic thought is a strategy for problem solving or decision making using mental shortcuts developed from previous experience. Heuristic thought arises unconsciously when people are confronted with complex or rapidly changing situations because attention, or time to reflect and analyze, are finite resources. Such thought processes are believed to have evolved from human origins as hunter-gatherers,3 so they apparently are ingrained. Thought biases, also known as cognitive biases, are examples of such mental shortcuts. Because they are based on prior successful experience, those mental shortcuts often result in accurate thinking and appropriate responses more efficiently than the “analyze everything” approach. In the pipeline industry, engineers, integrity managers, and other workers often have multiple responsibilities, work long hours under incredible time pressures, or must make decisions while subjected to information overload or uncertainty overload. In those circumstances, mental shortcuts made automatically are not just useful but necessary. But those shortcuts may also lead to errors that can adversely impact decisions4 including those that occur when managing the integrity of a pipeline or when performing other pipeline work. When information is incomplete or uncertain and the consequences of an incorrect decision are large, it becomes even more important to defend against these mental shortcuts. Numerous categories of thought bias have been described. A Wikipedia article5 lists no fewer than 220 named varieties under 21 categories, many with intriguing names such as the Google Effect,6 the Curse of Knowledge,7 the Hindsight Bias,8 the Hot Hand Fallacy,9 or the Women are Wonderful Bias.10,11 Almost all of the numerous biases, effects, or syndromes listed in the literature could influence decision making somewhere in a large organization, but it is more useful to focus on a few broad categories that occur more often with respect to managing the integrity of pipelines. These include: Cultural bias Availability bias 3 “16 Cognitive Biases that can Kill Your Decision Making”, www.boardofinnovation.com. Sunstein, C.R., “Moral Heuristics”, Behavioral and Brain Sciences, Cambridge University Press, 2004. 5 https://en.wikipedia.org/wiki/List_of_cognitive_biases 6 Sparrow, B., Liu, J., and Wegner D.M., “Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips”, Science, Vol. 333, August 5, 2011. 7 Kennedy, J., "Debiasing the Curse of Knowledge in Audit Judgment", The Accounting Review, v. 70, no. 2, April 1995 8 "I Knew It All Along…Didn't I?' – Understanding Hindsight Bias", APS Research News, Association for Psychological Science, Sept. 6, 2016. 9 Green, B., Zwiebel, J., "The Hot Hand Fallacy: Cognitive Mistakes or Equilibrium Adjustments? Evidence from Baseball", Stanford Graduate School of Business, April 2014. 10 Eagly, A.H.; Mladinic, A., "Are people prejudiced against women? Some answers from research on attitudes, gender stereotypes, and judgments of competence", European Review of Social Psychology, Mar. 4, 1994. 11 The authors acknowledge possible susceptibility to this bias. 4 4 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 Confirmation bias Representativeness bias The “flaw” of small numbers Sunk cost bias Groupthink Cultural bias Cultural bias is the interpreting or judging of matters by standards inherent to one’s group (in this case the pipeline company or a contractor). Cultural biases may discourage critical evaluation of long-held practices (“this is how we do it here”) as well as hostility to ideas or information originating outside the culture (the “not invented here” syndrome). Most of these biases appear valid based on experience (“Contractor/supplier X has always done good work”). But they are sometimes held without good evidence or even with contrary evidence, or they may be based on habits (“we always use that contractor/supplier” or “this has been our SOP for years”) that no longer stand the test of time as knowledge or technology changes. In the IM context, they may interfere with continual improvement, lead to a failure to recognize new hazards or risks, or directly lead to a poor decision. Availability bias Availability bias is a mental shortcut that focuses on the most recent information in one’s memory. The ease of recollection can lead to a false assumption that an event occurs more frequently than it does, causing one to focus on the wrong risk. Availability bias could occur due to specific problems on the most recent pipeline construction project, or the cause of the most recent incident, causing intense focus on avoiding those issues rather than a more general threat. This sometimes leads to company standards, technical procedures, or risk modeling covering a particular subject at a peculiarly high level of detail that is out of proportion to the coverage of other subjects. In the worst case it may lead to a complete distraction from another, potentially greater risk. Representativeness bias Representativeness bias occurs when one thinks a current situation is like another that may already exist in one’s mind. For example, “this is just like a situation we saw before”, or “we have done this many times in the past and it always worked fine” or more generally, the probability that event A belongs to group B based on how much A resembles B. While this seems like a logical way to judge things, it may lead to overlooking the unique aspects of the problem, underestimating the probability of a particular threat, or overestimating the effectiveness of a particular solution to a problem. A problem that arises from using this heuristic is the insensitivity to prior probability. In this case, people will judge the probability of an event purely based on how well something resembles one group or another regardless of the percentage of the population that truly belongs to that class. For instance, if a person is trying to decide if something belongs to Group 1 or Group 2 and decides that the description fits 80% of the population of Group 1, they will assume that the probability that it belongs to that group is 80%. Even if they are told that Group 1 only makes up 1% of the population, they will not change their initial probability judgement (the correct answer considering the prior probability is about 4% which is far below the assumed probability).12 These failures to thoroughly think through the matter have led to choosing incorrect mitigations or other actions. 12 Anderson, J., “Optimizing Risk Decisions with Imperfect Data”, Paper #36, PPIM2023, Houston, February 2023. 5 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 The “law” of small numbers The reason for putting the word “law” in quotation marks is that it’s not a mathematical law but a fallacy of faulty reasoning from insufficient evidence due to small sample size. People have an exaggerated faith in small samples because they want to assign causality. Yet the only “cause” might have simply been chance. For instance, 6 confirmation digs performed for an ILI run with 5 of them being within specification leads the engineer to conclude that the tool met the requirements for 80% within specification. After all, 5/6 = 83%, right? However, since the sample is so small, even if the tool was only performing to 50% within specification it would be expected to see that same result (5 out of 6 in-spec) by chance almost 10% of the time, not an insignificant probability that can be ignored. This goes hand in hand with the “flaw of averages” which states, “decisions based on averages are wrong on average”13. The tool performs to 80% within specification on average but that does not mean it performed to 80% within specification on this run. Confirmation bias Confirmation bias is selectively paying attention to information that agrees with what one believes to be true. It may lead to failure to recognize other important information. One example we observed involved an integrity engineer who was enthused about a recent in-line inspection (ILI) that had identified some significant hook cracks. But he had disregarded the dig data showing a large proportion of hook cracks observed in the ditch were not indicated by the tool and they were no less severe than the ones indicated by the tool. This suggested a high likelihood of other cracks existing undetected in unexamined joints of pipe. Can that ILI really be considered a success? Sunk cost fallacy The sunk cost fallacy occurs when one is reluctant to abandon a course of action because of having already heavily invested time or resources in it, even after it becomes clear that a different course of action would be more beneficial going forward. For example, we have observed this bias to cause operators to doggedly continue with a “pig and dig” approach to managing the integrity of a pipeline at great cost although analysis of the results showed that the only effective way to lower risk is to replace segments of the pipeline. Groupthink Groupthink occurs within a group of people in which the desire for harmony or conformity, usually driven by adherence to a defined process, results in dysfunctional decisions. It can cause a group to minimize conflict and reach a consensus without critical evaluation of the data or alternative actions, suppressing dissenting viewpoints (by punishing “rocking the boat”). It may also lead to an insular approach uninformed by outside ideas or experiences that could be effective or valuable. Several of the thought biases described above, and others, are noted in Section 7.3 “Cognitive limitations in decision making – Heuristics and biases” in CSA EXP24814. 13 Sam L. Savage, “The Flaw of Averages: Why We Underestimate Risk in the Face of Uncertainty” (Wiley, 2009) 14 Canadian Standards Association (CSA), “Pipeline Human Factors”, EXP248:2015. 6 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 Some Real Life Examples As stated earlier, bias is the making of unwarranted assumptions or making assumptions that are not based on demonstrated experience or accurate data. Bias is introduced in decisions within the IMP or in executing the IMP, as well as in routine work or in new construction activity. It causes us to fail to recognize threats or make the appropriate response. A few illustrative cases are described below. Bellingham, WA (1999): The Bellingham, WA incident occurred at mechanical damage caused during installation of water treatment plant piping over the pipeline. The ILI tool identified the anomaly that eventually ruptured but because the site was difficult to access and after examining “several” other anomalies, the person in charge concluded that the tool was over-reporting the actual depth. So, a decision was made that the anomalies near the water treatment plant would not be excavated and examined.15 The probability of a 4% dent being an integrity threat given that the tool over-called other deformations is low. However, the probability of a 4% dent being an integrity threat given that it was not called on the previous run and that the location coincided with where a large water line had been installed over the pipeline since the last run is very high. The investigation also revealed that despite many meetings with the contractor working at the water treatment plant and site visits by the operator, the operator did not have full awareness of work at the site. Last minute design changes were not communicated, and a witness reported that the excavator chose not to report hitting the line. Compounding these conditions, the operator had been updating its SCADA system without validating the changes offline. These interfered with the ability of controllers to respond effectively to the pipeline failure. The biases that contributed to this incident included: Cultural bias – Overconfidence in weak administrative barriers against excavator damage, inadequate management of change for the software update Representativeness bias – the unexamined ILI feature would be like those that were examined Confirmation bias – Disregard of the fact that the unexamined dent was a new indication at a recent work site Law of small numbers fallacy – Overreliance on conclusions from a small number of observations. Carlsbad, NM (2000): The pipeline failed due to internal corrosion. The operator assumed that an existing drip would catch any liquids in the line and that warning signs and the remoteness of the site would keep people out of the area even though the affected site was on property not under the operator’s control. 15 NTSB, “Pipeline Rupture and Subsequent Fire in Bellingham, Washington June 10, 1999”, NTSB/PAR0202, PB2002-916502, https://www.ntsb.gov/investigations/AccidentReports/Reports/PAR0202.pdf. 7 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 However, per the NTSB report16, the drip siphon had become plugged. When the collection leg filled up liquids were pushed downstream to the next low point. The drip stopped producing liquids when it was checked which seemed to confirm a reasonable but incorrect assumption that there were no liquids in the line. The pipeline operator routinely collected liquid samples from inlet scrubbers and maintenance pigging and had them analyzed for environmentally hazardous substances but never tested them to determine if they were potentially corrosive, even though an internal corrosion failure had previously occurred elsewhere in the system and the segment that failed was incapable of being pigged. The biases that contributed to this incident included: Cultural bias – Failure to recognize an internal corrosion problem, not collecting the data needed to manage the condition, and overconfidence in weak barriers against failure (the drip) or consequences (warning signs) Confirmation bias -- The drip was not producing liquids so they must not be present Representativeness bias – What they had been doing was working so far so it was adequate. A Series of Unfortunate Events17 The two incidents cited above were selected specifically because they led to integrity management planning as specified in US pipeline safety regulations. Thought bias clearly played prominent roles. But integrity management regulations and industry practices have not eliminated the role of thought bias in incidents. The authors believe that almost any incident or mishap can be analyzed in terms of unwarranted assumptions stemming from some type of bias. Some further examples follow. RSI personnel have performed root cause failure analyses (RCFAs) of several incidents and near misses in a pipeline operator’s systems. The operator was technically proficient, had detailed procedures in place, and the incidents seemed unique and completely unrelated. But a common thread running through all of them was cultural bias that promoted overconfidence in the effectiveness of barriers against failure including threat identification, material specifications, field practices and procedures, construction inspections, in-line inspection, and organizational communication, among other things. An NTSB study18 from 2005 found that a SCADA issue played a role in 10 of 13 hazardous liquid pipeline incidents. In 7 of those 10 cases, an inability by controllers to distinguish between alarms triggered by spurious or normal event conditions and those triggered by real emergencies contributed to delayed or incorrect actions that caused or increased the severity of the events. The controllers often made incorrect assumptions about the causes of the alarms based on many prior observations of false alarms associated with other common operating situations, which were indicative of classic representativeness bias. Although these incidents preceded control room management requirements in pipeline safety regulations, RSI’s RCFA team have observed false-alarm overload in contemporary events. 16 NTSB, “Natural Gas Pipeline Rupture and Fire Near Carlsbad, New Mexico August 19, 2000”, NTSB/PAR-03/01, PB2003-916501, https://www.ntsb.gov/investigations/AccidentReports/Reports/PAR0301.pdf. 17 This is not a reference to the similarly titled children’s book written by Daniel Handler under the pseudonym Lemony Snicket. 18 NTSB, “Supervisory Control and Data Acquisition (SCADA) in Liquid Pipelines”, Safety Study NTSB/SS05/02, PB2005-917005, Notation 7505A, November 29, 2005. 8 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 RSI personnel have observed numerous incidents involving defective new pipeline construction owing to unjustified confidence in the competence or good intentions of material suppliers, construction contractors, fabricators, welding inspectors, site inspectors, or other parties. Positive or negative change in a contractor’s business, or change in ownership, has been observed to adversely affect performance in some cases. The unjustified confidence led to inadequate supervision and control of construction quality at differing stages in many projects. Most of these led to damage-only outcomes but a couple events led to injury or fatality. Those that did not have such serious consequences had the potential for worse and that they did not was a matter of luck.19 With respect to this last item, the authors recognize the need to trust that people can and will do the job they are tasked with and do it well. In fact, most people intend to. But time and again, operators are too quick to conclude that once qualified, it is no longer necessary to think about that supplier, contractor, or worker, or that once written, it is no longer necessary to think about the effectiveness of specifications and procedures. We have often observed that the engineers in the glass towers downtown have no idea what is happening in the field. The Cloak of Invisibility Pipeline safety and integrity management regulations and industry guidance documents are insensitive to patterns of thought bias that may undermine their goals. They may even promote the making of unjustified assumptions primarily by error of omission. Some examples follow: The “TTO5” baseline seam assessment decision chart20 applies to pressure-cycle fatigue of low-frequency ERW seams. Some operators have incorrectly assumed that since highfrequency ERW21 and DSAW seams are omitted they are not susceptible to fatigue. ASME B31.8S22 and Part 192, Subpart O both state that the manufacturing defect integrity threat is mitigated in a natural gas pipeline if it has sustained a hydrostatic pressure test to at least 1.25 times the MAOP. However, studies have shown that a test pressure ratio of 1.25 is effective only for pipelines operating at Class 1 stress levels, but it is inadequate for pipelines operating at lower stress levels. ASME B31.8S states that pipelines that operate at stress levels of 60% of SMYS or greater are susceptible to stress-corrosion cracking (SCC). Many operators have then incorrectly inferred that pipelines operating below that stress level are not susceptible.23 The pipeline regulations specify minimum criteria for CP without limiting overprotection. To comply, some operators overprotect their pipelines in an attempt to meet the minimum requirement over all parts of a segment, potentially disbonding coatings or inducing a risk 19 They were what are sometimes referred to as “near-miss” events. Michael Baker Jr., Inc., Kiefner and Associates, Inc., and CorrMet Engineering Services, PC, “Low Frequency ERW and Lap Welded Longitudinal Seam Evaluation”, Report to US DOT, RSPA, PHMSA, Delivery Order DTRS56-02-D-70036, Final Report, Rev. 3, April 2004. 21 Kiefner, J.F., and Kolovich, K.M., “ERW and Flash Weld Seam Failures”, Subtask 1.4, US DOT, DTPH5611-T-00003, September 24, 2012, compiled failures in ERW and flash welded seams due to various causes. Five of the 37 cases reported as due to fatigue crack growth occurred in high-frequency ERW seams. The RSI RCFA team is aware of others since the publication of that study. 22 ASME, “Managing System Integrity of Natural Gas Pipelines, Supplement to B31.8”, B31.8S-2020. 23 This is like stating that driving 90 mph could be considered speeding, which is true, but driving well below 90 mph could also be considered speeding in many circumstances. 20 9 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 of hydrogen cracking, instead of taking steps necessary to manage CP to more uniform levels that are not excessive. The criteria or guidance in these examples are correct as far as they go, but they are incomplete statements of the concern through error by omission. On the other hand, it is impossible to write enough rules to prevent all errors or to account for all variations of complex problems. Operators are responsible for correctly understanding the intent of the provisions, and for applying appropriate technical knowledge to their interpretation; many operators have done so successfully. Those that made unjustified assumptions or convenient interpretations have sometimes discovered their errors through unhappy experiences. For the most part, the regulations and guidance documents are not written in a way that promotes discovering erroneous assumptions, though exceptions exist. For example, ASME B31.8S requires annual review of risk model assessment assumptions. It also requires performing evaluations of timedependent conditions using appropriate defect growth rates to assure that a failure will not occur before the next integrity assessment. The guidance for the “check” portion of the Deming Cycle described in API 116024 recommends activities having the potential for discovering and correcting bias. Both Parts 192 and 195 require that ILI be performed in accordance with API 116325 which provides methods for validating the ILI tool performance. The regulations and guidance documents all require evaluating IMP effectiveness, though that does not assure discovering built-in bias if the performance metrics are biased. API 1160 provides recommendations for self-review that can be helpful in that regard. However, procedures that are written as compliance documents will not inherently reveal bias written into them or in their implementation., nor will internal audits that are box-checking compliance exercises. Recognizing and Preventing Bias Assumptions by themselves are not deleterious to the decision process. They become a fallacy when the decision maker fixates on an assumption and refuses to move from it. A rigorous decision process allows for the data to change the opinion even if it goes against the prior assumptions. The decision process needs to assess all the information that is available not just the parts that fit the initial assumptions. Question your beliefs Awareness of bias in your own thought process is a place to start.26,27 As you consider your decisions about threat identification, ILI tool selection, contractor selection, or almost anything else, challenge your own perception of the problem. Ask yourself questions such as: 24 American Petroleum Institute, “Managing System Integrity for Hazardous Liquid Pipelines”, Recommended Practice 1160, 3rd Ed., February 2019. 25 American Petroleum Institute, “In-line Inspection System Qualification”, Standard 1163, 2nd Ed., 2013, 2018. 26 We refrain from saying “knowing is half the battle” as that is a classic biased response known as the “G.I. Joe Fallacy”, named after a 1980s TV show in which every episode ended with a short object lesson and the adage “Now you know. And knowing is half the battle”. To the contrary, it has been shown (Kristal and Santos) that certain biases that are embedded emotionally cannot be overcome by awareness and conscious reflection alone. In those cases, knowing is much less than half the battle. 27 Kristal, A.S. and Santos, L.R., “G.I. Joe Phenomena: Understanding the Limits of Metacognitive Awareness on Debiasing”, Harvard Business School, Working Paper 21-084, 2021. 10 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 Why do I think this? Is this always true? Am I jumping to conclusions? What am I missing? Is this what the data shows? Is my data reliable? Do I believe this inspection report? Is this ILI feature what it is reported to be? Your questioning can extend to the beliefs you hold as an organization: Why do we do it this way? Is it still the best way to do it? What do other operators do? We have procedures for everything, but are they being followed? We get all these reports from the field but what are they telling us? Do we even look at them? We have been using this contractor/suppler/consultant for years, but are they still the best choice? Have we recently evaluated their performance? The answers may be unsettling. If that is the case, you may have encountered a personal view or an organizational behavior that has been influenced by an unproductive bias. Behavior of a high-performance organization When driving a vehicle in challenging conditions (heavy or fast-moving traffic, bad weather, winding or poorly lit roads), the driver must continuously perform at a high level of attention and skill. A momentary lapse in attention, or a deficit in skill or judgment, may cause the driver to fail to respond to a situation with the near-instantaneous correct judgment and coordinated response needed, which can be catastrophic.28 Vehicles are routinely operated in unforgiving environments that present the potential for error and risk to others. Pipelines also operate in unforgiving environments (physical, social, and political) and present the potential for error and risk to the public.29 Managing the risks on the road requires the driver to perform at a high level; managing the risks of a pipeline requires the operator’s organization to perform at a high level. Business consultancies, educational institutes, and publications that cater to the business market often try to distill the distinguishing characteristics of a high performance organization (HPO) to a Top 5 list, such as: communication (and buy-in) of values, reinforcement of positive behavior, open communication, management trust of employees, and collecting feedback.30 Others list safety performance and skill development31 or accountability and flat organizational structures.32 Any list of 5 characteristics is inadequate. Some characteristics at the staff and management level of a HPO might include:33 28 According to the National Safety Council, in 2020 there were over 42,338 motor vehicle fatalities. Most Americans are exposed to hazards from automobiles either as drivers, passengers, or pedestrians. Based on 2021 data from various sources, the authors estimate that approximately 50% of the US population is exposed to the risk from natural gas distribution, 15% from natural gas transmission, and 11% from hazardous liquid pipelines. With 10-year average annual fatalities rates of 2.1, 1.4, and 7.7, respectively, if there were enough pipelines to expose 100% of the US population to the risk, the projected number of fatalities would be around 42/yr. Thus, automobiles pose an average 42,000/42 = 1,000 times greater net societal risk than do pipelines, but pipelines do pose a larger potential hazard on an incidental basis. 30 https://www.ottawa.edu/online-and-evening/blog/november-2020/5-key-elements-to-creating-a-highperformance-comp 31 https://www.industryweek.com/leadership/article/21146834/what-makes-a-highperformance-organization 32 https://www.bcg.com/publications/2011/high-performance-organizations-secrets-of-success 33 Discussions with Dr. Alan Murray, Pipeline Consultant. 29 11 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 They do what they say they will do They have the courage to do what is right They hold themselves and others accountable They understand what they are doing and why They anticipate problems and are alert for unusual conditions They ask questions, consult, and verify, rather than act on assumptions They value collective input and the views of others They speak up when problems are identified They communicate information in a disciplined way They seek staff development training They are willing to listen to contrary opinions It is more difficult for bias to thrive and persist in such a culture. The authors have observed organizations in which management holds the above values to be important, but either that message is somehow not communicated convincingly to staff, or the intended behaviors become overwhelmed by time pressures or attention deficits that cause individuals to go into a “survival mode” in which they fall back on heuristic patterns in the moment of decision. On the other hand, there are many cultural behaviors that clearly will promote thought bias and other unproductive behaviors in many forms. Some of those include: Management by intimidation Starving integrity budgets Overworked staff Impossible deadlines Management inability or unwillingness to either motivate or get rid of continually poor performing employees, which demotivates high performing workers Intracompany competition for resources Subject matter expert arrogance or dismissiveness Overconfidence in procedures, knowledge, and barriers against failures Denial or trivialization of errors Information hoarding Failure to hand off records – Construction to O&M, O&M to Integrity Siloing – a lack of communication or information sharing between groups Lack of constructive or relevant training or professional development Echo chamber – low awareness of problems, developments, or trends in the industry Unwillingness to look at new ways of doing something (“not invented here”) Reliance on rules of thumb Procedures and integrity plans written to comply with the letter of regulations 12 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 We are not alone Civil engineers have attempted to classify the causes of structural failures much as the pipeline industry has done. In one study34, the main factors which affect “proneness to structural accidents” were: 1. 2. 3. 4. 5. 6. 7. 8. New or unusual materials New or unusual methods of construction New or unusual types of structure Experience and organization of design and construction teams R&D background Financial climate Industrial climate Political climate The fourth leading cause was related to organizational issues. Another study35 that focused on failures of steel bridges concluded that 10 of 24 cases were the result of human error in the design or construction phase. A Canadian study36 confessed that the construction industry (at that time) lacked an understanding of the basic factors involved in human error in design and construction of structures and suggested turning to other disciplines such as human behavior for guidance. The Chernobyl disaster in 1986 shocked the international nuclear power industry into developing the concept of the “safety culture” to explain the organizational dynamics that led to the accident.37 A definition of safety culture was developed by the British Advisory Committee for Safety in Nuclear Installations (ACSNI)38 which was generalized to all hazardous industries: “Safety culture is the product of individual and group values, attitudes, competencies, and patterns of behavior that determine the commitment to, and the style and proficiency of an organization’s health and safety programs. Organizations with a positive safety culture are characterized by communications founded on mutual trust, by shared perceptions of the importance of safety and by confidence in the efficacy of preventive measures.” This was 10 or more years after the Canadian structural industry contemplated delving into behavioral science to explain structural failures. The pipeline industry has moved tentatively in that direction also with the concept of a pipeline safety management system (PSMS).39 The PSMS sets 34 Pugsley, A.G., “The Prediction of Proneness to Structural Accidents”, The Structural Engineer, 51(6), June 1973. 35 Blockley, D.I., “Analysis of Structural Failures”, Proc. Inst. Civil Engineers, January 1977. 36 Allen, D.E., “Structural Failures Due to Human Error: What Research to Do?”, National Research Council Canada, Proc. Symposium on Structural Technology and Risk, July 1983. 37 Wilpert, B. and Fahlbruch, B., “Safety Culture: Analysis and Intervention”, Conference Proceedings, Probabilistic Safety Assessment and Management, 2004. 38 Advisory Committee on the Safety of Nuclear Installations, Study Group on Human Factors (1993), Third Report: Organizing for safety, London: HMSO, 1993. 39 American Petroleum Institute, “Pipeline Safety Management Systems”, Recommended Practice 1173, July 2015. 13 Pipeline Pigging and Integrity Management Conference, Houston, February 2023 forth requirements for leadership, middle management, and employees that, as it turns out, mirror many of the favorable behaviors listed earlier for an HPO but directed toward safety. Thus, developing and maturing a PSMS may eventually contribute to reducing the risk of thought bias interfering with the effectiveness of an operator’s integrity management plan. Though a PSMS system, even if carried out correctly, will not debias any individual, it does have the potential to make the organization less reliant on bias. Acknowledgements The authors wish to thank RSI’s RCFA team colleagues Ms. Stephanie Flamberg and Ms. Cara Macrory for their insightful observations in many complex RCFA investigations. The authors also wish to thank Dr. Alan Murray for his astute comments over many discussions about the role of bias in incidents. 14