Causal Learning Overview

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Purpose of Causal Learning
• Despite our best efforts we continue to experience
undesirable outcomes such as:
• Injuries
• Releases
• Production losses
• Cost overruns
• Performance Gaps
• Causal Learning provides the opportunity for us to
learn from the causes of these undesirable
outcomes and create the performance we want
instead.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Purpose of Causal Learning
• Causal Learning helps us improve performance
by:
– Discovering the causes of performance problems
– Learning from the discovered causes
– Implementing solutions that address causes and create
the desired performance instead

• Causal Learning is used to solve problems and to

investigate incidents in all areas of performance
– HSE, Production, Quality, Projects, Financial, etc.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 How performance is caused

Outcomes
– Physical causes/mechanism

Human behaviour
– Decisions and actions

System Leaders
– Structure, resources, create the
expectations, consequences system

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Key Elements of Causal Learning
Causal Learning upgrades the capacity of the
organization to improve performance through
• Causal reasoning - Discovering the causes of our
current performance
• System thinking - Seeing the underlying causal
patterns across performance outcomes that we
have experienced
• Learning - Shifting our understanding of how
undesired outcomes are created
• Corrective action - Taking action that addresses
the causes of undesirable performance and
creates the performance we want instead

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Causal Learning Core Concepts

• Every effect is caused Defense Solution

• Causality is positive

• Data has quality

• Outcomes are caused by the
performance system

• Learning changes Action

• Corrective action addresses causes

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
Obstacles to Performance Improvement

• Action Priority – We seek to fix a problem

before we understand what caused it
• Generalization – We try to address
problems in general terms without looking
at the specific causes
• Defensive Reasoning – We see a problem
through what was not done rather than
what actually happened
• Hindsight Bias – We see the problem
through knowledge of the outcome
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Hindsight Bias

• People make choices to achieve a desired outcome Goal

based on the situation they are in and their beliefs
at that time
• After an unwanted outcome occurs instead, we can
look back at the outcome and see what they should
have done differently
• We can then judge the people involved for making
choices that we believe (now) they should not have
made (then)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Reasoning Orientations

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Reasoning Orientations
Causal reasoning:
Determining what actually
happened that created the problem

Defensive reasoning:
Determining what failed, was
missed or was wrong

Solution Defense
Solution reasoning:
Determining what action to take to
fix or prevent the problem

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Incident Example 1
A flange gasket on a flow control valve in a line containing
caustic liquid failed, leading to a release of the caustic
material. An operator needed to take a routine water sample
in the same area. He was walking past the line at the
precise time that the gasket failed. The valve was mounted
vertically in a line at about five feet above grade, putting it
at eye level for the operator. The spray of liquid from the
valve when the gasket failed struck the operator in the face.
The operator was rushed to an eye wash area by a fellow
worker and then taken by ambulance to the hospital
afterwards. The injury was a recordable incident.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 The operator did not The operator was in the area
recognize the potential The operator did not wear
goggles. to take a water sample.
hazard of that valve.
Was the right gasket What was the operating We need to reinforce the
installed? pressure in the line? procedures for wearing eye
The flange was located at protection.
We need to instruct
eye level. Move the water sample
maintenance on the right
Had the gasket been station away from the
gaskets for that service.
replaced recently? caustic area.

Causal Defensive Solution

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Reasoning Orientation Review

• How can you tell when you are using

each reasoning orientation?
• In which reasoning orientation do
you feel more comfortable?
– Why?
• When would each reasoning
orientation seem more appropriate?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Reasoning – Key Points
• Defensive Reasoning
– Triggered by hindsight bias (what it looks like after we know the
outcome)
– Focused on “what they should have done differently” and
“what failed or did not happen”
– Reinforces existing barriers and controls, based on what we
already know or believe.
• Causal Reasoning
– Uses foresight perspective (what it looked like before the
outcome)
– Focused on why it actually happened
– Stimulates learning and allows action to address causes
• Solution Reasoning
– Focuses on action to “get out of the problem”
– Actions are based on assumed causes.
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Overview of
Causal Learning
Process

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning - Sequence
• Sponsor for Performance Improvement
– Clarifying purpose, scope, resources, outcomes
• Conduct the investigation
– Discover the causes, engage with Sponsor, decide when done
• Conduct Learning Session(s) with others
– Provide opportunity to learn and to broaden Insights
• Decide what to address
– Managers select causes to address
• Develop corrective recommendations
– Develop effective solutions to address selected causes
• Action Commitment with Managers
– Gain alignment on recommendations to implement
• Develop and implement solutions
– Managers lead solution implementation
– Check that performance actually improves

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Sponsor Decision to Investigate
• Impact of incident or problem (actual or potential)
– “Must” investigate

• Perceived as “repeat” incident

– Problem remained despite our efforts to address it

• Opportunity for learning seems high

– Don’t understand why it happened, but think we should
– Believe it never should have happened and realize there are
aspects of the system we don’t understand

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Responsibilities of a Sponsor
• Defining the purpose and scope of the effort
• Demonstrating the importance of the work
• Selecting the Causal Learning facilitator and team
• Helping develop Terms of Reference
• Staying in touch throughout investigation
• Ensuring access to resources and helping remove
barriers
• Ensuring investigation is discovering what is needed
• Leading the learning sessions and decision making
meetings
• Supporting and sustaining the work through solution
implementation

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
Causal Learning Investigation Team Work
• Team Formation basis the TOR
• Data Gathering
• Analyzing data using CL tools
– Problem Statement
– Timeline
– Cause and Effect Diagram
• Learning session(s)
• Corrective recommendations
• Written report

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Investigation Roles
• Sponsor
– Creates conditions for success
• Facilitator
– Provides/facilitates the investigation process
• Team Leader
– Team member and link to sponsor
• progress, resources, etc.
• Team Member
– Learns and uses the process to conduct the
investigation per the TOR

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Data Gathering

Interviews

Observations Records
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Investigation Method

Observations
Problem Statement Insights Timeline
Conclusions

Cause and Effect

Diagram

Is - Is Not Data Quality

Corrective
Recommendations

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Investigation Tools
• Problem Statement – Difference between what
was expected and what actually happened
• Timeline – Relationship of events over time
• Is - Is Not – Difference between what happened
versus what could have, but did not happen
• Cause & Effect Diagram – Analysis of cause-
effect relationships
• Data Quality – Level of certainty that the causes
are true
• OIC's – Sharing what we are learning
• Corrective Recommendations – Suggested
action to eliminate causes
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement - Example
• Expected:
During the turnaround, some unexpected work will be
discovered after equipment is opened up and
inspected. The discovery work is normally small enough
that the overall turnaround schedule can be maintained.
• Actual:
Large amounts of coke were found in the reactor system
plus erosion of many tray valves that was not anticipated
in the turnaround scope.
• Impact:
The turnaround schedule was delayed to remove the coke
and make repairs. During coke removal additional
refractory was damaged leading to further repairs.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Timeline

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Is - Is Not Analysis Format
Is Is Not Difference
What

Where

When

Magnitude

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Diagram

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Data Quality “Ladder”
Fact - Verifiable, precise, measurable
Inference - Logical deduction from fact
Hypothesis - Causal theory that, if true,
would explain the facts
Assumption - Perception derived from one’s
own experience
Common belief - Perception derived from
experience of others
Hearsay - Perception based on no experience
(2nd or 3rd hand information)
Guess - No factual or logical basis required

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
Observations, Insights, Conclusions
• Observation:
– The data that “struck” me the most (My
“evidence” with no assessment)

• Insight:
– The meaning I make from my observation
(my “aha”)

• Conclusion:
– If (my insight) is true, then this (my
conclusion) might also be true
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Development
Process
1. Decide what cause to address
– What you want to change about the existing system
2. Describe what you want to create instead
– Achieve, avoid, maintain
3. Develop Solution Ideas
4. Test Solution Ideas for Effectiveness
– If the solution had already been in place, how would it
have prevented the incident and what would have been
created instead?
– How effective will the solution be to prevent similar
incidents in the future? Could the solution create
unintended consequences?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem
Statement

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement
Analyzes a problem through difference
between what actually happened versus
what was expected to happen

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement Benefits
Defines the focus of the investigation:
– Helps identify and agree with sponsor
the specific problem to be investigated
– Helps distinguish among multiple
possible problems
– Clarifies differences between
expectations and the actual event
– Focuses investigation on discovering
the causes of what actually happened

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement Format
• Expected:
– What we expected to happen
(before the incident occurred)
• Actual:
– What actually happened
• Impact:
– Actual or potential consequences of
what happened

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement Example
• Workers bolting flanges at height
• Precautions taken in tool handling
• Pry bar slipped through grating, struck worker below

• Expected: What did they

expect to happen?
• Actual: What actually
happened instead?
• Impact: Actual or
potential consequences

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Problem Statement Tips
• The “expected” is what people would have
said they expected to happen in this
situation if you had asked them prior to
the incident
– It is not what “should have” happened or what
you desire to happen in the future.
• The “actual” is what did happen during the
incident
– Not what “did not” happen.
• It may be easiest to describe the “actual”
first
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Interviewing
Skills

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Deciding who to interview
Interview candidates
• Witnesses to incident
• Any one with a role in the incident or
events leading up to incident
• Others in similar roles
• Manager of involved personnel, possibly
at several levels
• Expert on relevant systems or technology

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Interviewing Approaches
Timeline approach (What)
• What happened (sequence of events) from time “X”?
• Best for participants or witnesses to incident
Technical approach (How)
• How does that (equipment, process) work?
• Best for experts on a relevant area
Focused approach (Why)
• Used to test specific causal theories
• Usually need a set of selected questions

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Some Interviewing Tips
• Be clear as a team what you are trying to find out
(Before interview)
• One interviewee
• One or two interviewers (as appropriate)
• Get acquainted first (Don’t jump right into questions)
• Explain why you are there
• Avoid leading questions (ones that suggest an
answer) and yes/no questions
• Don’t ask “why”. Ask “what” or “how” questions.
• At end, ask: “Is there anything else” and “Can we
come back later?”
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Timeline

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Timeline
Timeline analyzes a problem through the
lens of relationships of events in time

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 What is a Timeline?
• Timeline organizes events on a consistent
linear scale
• Timeline is an analytical tool that focuses
on sequence (not causality)
• Looking at your data on a graphical
timeline allows you to develop insights
from “what happened”

1991 1992 1993 1994 1995 1996 1997 1998 1999

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 How to develop a Timeline
• Show all actual “events” gathered from the
investigation in sequence on the timeline

Documents
Interview
Data Computer
data

1991 1992 1993 1994 1995 1996 1997 1998 1999

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Timeline Options
• Timelines can vary in design but time
scale should be consistent

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Timeline Development Tips
• Show "what happened" not "what should or could have happened"
• Add events you think may be relevant such as:
– When problem occurred
– Events before or during problem
– When problem became bigger or smaller
– Events related to possible causes
– Actions taken to address the problem
• Keep a consistent time scale. You may need to create separate
timelines as data shifts from months to days to minutes
• If you don’t know the exact time for events, put them down in the
sequence you believe is true

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and
Effect
Analysis

Possible
Cause

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Analysis
Cause and Effect analysis views the
problem through cause and effect
relationships.
Apply causal reasoning in three phases:
• Generating possible causes
• Tightening cause and effect logic
• Verifying causes with data
Output: Cause and effect diagram.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
What is a Cause and Effect Diagram?

A graphical representation of cause and effect logic

• Logical relationships between boxes: “and” “or”
• One causal statement per box
• Causes are actions or conditions

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Actions or Conditions
I got wet

And
It was raining I was walking
outside outside

I got wet I got wet

And And
It was I went I was It started
raining outside walking to rain
outside outside

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Writing the Top Box on Diagram
• Connect to Problem Statement
– What is the “problem” you want to A

understand? B

– At what causal level does selected

“problem” occur (A or B)?
• Connect to Timeline
– At what moment in time did the
top box occur?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Logic

To identify possible causes, ask: Effect

• Why did this happen?

• What would have to be true for this to happen?

• What else could cause this to happen?

Cause

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Possible Cause Generation
Purpose: Create what possibly could have
caused the problem
Characteristics
• Go where logic takes you
• Accept uncertainty
• Accept it will change later
• Explore

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Possible Cause Generation
Tips:
• Focus on what happened (not what should have happened)
• Consider one effect at a time
• Develop one level at a time
• Write complete sentences (specific thoughts) in each box
• Be complete. Causes need to explain:
– Location
– Time
– Magnitude
• Use past tense. Causes occurred in the past.
• Take small cause/effect steps
• When stuck, clarify “How we believe this happened (our story)”
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Possible Cause Generation Tip
To discover Why something
happened, you will first need a
common understanding of How it
happened.
• Use visual aids
• Create Drawings
Story 1
• Write Stories
Story 2

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and
Effect
Analysis

Human
Behaviour

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Determining Causes of Human
Behaviour - Principles
• Human behaviour is caused
– The causes of human behavior can be discovered using
causal analysis
• People behave based on what they know or have the
capacity to do
– Not on what they don’t know or don’t have the capacity to do

• Behaviour is influenced by the system in which the

individual works
– Cause and effect analysis must examine the “system”, not
just focus on the individual

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Human Behaviour Model
Feedback

Reasoning

Resources Consequences

Observations Actions
Goal
Structure Expectations

Knowledge/Beliefs
Experience

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Possible Causes of Actions/Decisions
• Reasoning
– What was perception of situation?
– What was Desired Goal?
– What Experience/Knowledge did person have?
– How would this action achieve something? (Logic)
• Resources
– What tools were available?
– What information was provided?
• Consequences
– What was perceived (positive or negative)?
• Structure
– What were job requirements?
– What role, policies, procedures were relevant?
• Expectations
– What was expected? (group norms)
– What were recognized practices?
• Beliefs
– What happened before (experience)
– Meaning made from past experiences (learning)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Observations,
Insights and
Conclusions

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Modes of Learning
 Operational (Psychomotor) Learning – adds new physical
skills (which tend to be permanent or easily reactivated);
 Informational (Cognitive) Learning – involves knowledge,
comprehension and critical thinking (which may be applied
depending on perceived value or frequency of use);
 Transformational (Affective) Learning – shifts, or develops new
values or beliefs (which influence what a person does and their
behaviours)
Transformational learning – personal, fleeting, uncomfortable
Increase the likelihood by providing good:
• Conditions for learning (e.g. leadership, environment, time)
• Input (e.g. timelines, causal charts, views of others)
• Structure (e.g. OIC’s, My Part)
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Learning is about…
• Personal learning
– Developing new insights from your
experiences
– Deepening your understanding
– Shifting your perspective
• Collective learning
– Sharing your personal perspective
– Deepening our collective understanding

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
Observations, Insights, Conclusions
• Observation:
– The data that “struck” me the most (My
“evidence” with no assessment)

• Insight:
– The meaning I make from my observation
(my “aha”)

• Conclusion:
– If (my insight) is true, then this (my
conclusion) might also be true
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 OIC Tips

• Make sure to include your observation

• Avoid stating what should have been but
was missing (Defensive Reasoning)
– “did not”, “insufficient”, “inadequate”, etc.
• Avoid stating your conclusion as an
action (Solution Reasoning)
– “We need to”, “we should”, etc.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 OIC Assignment
• Reflect on what you have observed so far
in this workshop
• Individually, write down your learnings in
the format:
– Observation
– Insight
– Conclusion
• Transfer one of your OIC’s to a flip chart
• Be prepared to share with others

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Using OIC’s
Investigation Team Learning

• Participants: Investigation team members

• Agenda:
– Reflect on investigation findings – every morning
– Conduct OIC activity with participants
– Discuss new insights and how this affects investigation

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Using OIC’s
Learning Sessions
• Participants: Individuals who can learn from (see
themselves in) the causes and can help shift the
performance
• Agenda:
– Present what was discovered
• Incident Description
• Problem Statement
• Timeline
• Cause and Effect Diagram
– Conduct OIC activity with participants
• Set up OIC exercise
– Discuss forward path (not solutions)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Learning– Key Points
• Shifting what people want to do / how they behave
involves Transformational (Affective) learning
• To be effective this learning needs:
– A good understanding of the causes of our performance
problems (i.e. causally reasoned)
– The appropriate conditions / environment
– To be sustained / supported until embedded
• The OIC tool is a powerful way to stimulate
personal, and collective, Transformational
learning
• Time allocated by the investigative team for their
learning helps both the work of the team and their
ability to support the learning by others
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Team
Facilitation

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Team Facilitation Assignment
Individually
• Reflect on your personal behaviors and the
behaviors of others on your team
– What behaviors have been helpful to your team?
– What behaviors have been less helpful?
As a team
• Share/discuss your observations with the team
• Decide on what adjustments to make
• Apply those adjustments when you continue
working on your causal chart

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and
Effect
Analysis

Logic
Tightening

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Analysis
Cause and Effect analysis views the
problem through cause and effect
relationships.
Apply causal reasoning in three phases:
• Generating possible causes
• Tightening cause and effect logic
• Verifying causes with data
Output: Cause and effect diagram.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
Cause and Effect Logic “Tightening”
Purpose:
– Ensure the logic is causal.
– Move from what “we think we know” to
what “must be true” logically.
– See what else may be causal.

Outcome:
– A cause effect diagram with accurate
and complete logic.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Logic
To test logic tightness, ask:
Effect
• How does this cause create that effect?
• Does this directly cause that?
• Are causes complete in terms of time, place
and magnitude?
• Is this cause necessary for that effect?
• Is the combination of these causes sufficient
to create that effect every time?
Cause
• Are we clear about actions and conditions?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Logic Tightening

Effect

Sufficient to AND AND

cause the effect
every time?
Cause Cause Cause

Necessary to
cause the effect?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Necessary?

Will …. ?
Effect

AND AND

Cause Cause Cause

If…

Necessary to
cause the effect?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Sufficient?

Will …. ?
Effect

AND AND

Every time…..
Cause Cause Cause

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Logic Tightening Tips
Avoid Example Alternative
Negative ”Eric did not recognize the “Eric thought it was safe to add
statements hazard" another neoprene wrap”
“The Neoprene patch remained
”The temporary repair
on the oily water drain line at
Generalizations stayed on the line for a
the 50 meter level until the 5th
long time"
of December.”

Judgments ”Eric was incompetent” What Eric did

Stating data as ”NRV 17 flapper was found ”The NRV 17 flapper had fallen
cause lying inside the body." off"
"There was an path to walk
Stating the absence "No safety barrier was
underneath the work platform
of a solution installed"
above"
"corrosion", "fouling", "lack Describe the specific
Vague statements
of lubrication" mechanism, location, timing
Copyright, all rights reserved,
Causal Learning
2020 Shell Exploration & Production Co.
 Cause and
Effect
Analysis

Data
Quality

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Cause and Effect Analysis
Cause and Effect analysis views the
problem through cause and effect
relationships
Apply causal reasoning in three phases:
• Generating possible causes
• Tightening cause and effect logic
• Verifying causes with data
Output: Cause and effect diagram

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 What is Data Quality?
• Data Quality is an assessment of the
investigation team’s perceived certainty of
a causal statement based on the quality of
their data and reasoning
• The Data Quality ladder provides a set of
levels (fact, inference, etc.) to assess the
investigation team’s perceived certainty of
a causal statement
• Assigning data quality is a key step in
cause verification

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Data Quality “Ladder”
Fact - Verifiable, precise, measurable
Inference - Logical deduction from fact
Hypothesis - Causal theory that, if true,
would explain the facts
Assumption - Perception derived from one’s
own experience
Common belief - Perception derived from
experience of others
Hearsay - Perception based on no experience
(2nd or 3rd hand information)
Guess - No factual or logical basis required

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Benefits of Data Quality
Assigning data quality:
– Tells team what data they have to support
their analysis and where they need to do more
research
– Provides a means of deciding when team is
“certain enough” about a causal statement
– Provides a means for team to communicate
their certainty about each causal statement

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Verifying Causes with Data
For each causal statement on your
diagram, ask:
1. What data do we have about this
cause?
2. Based on that data and our reasoning,
do we believe this cause is true or not
true?
3. How certain are we? (Using the data
quality ladder)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Sponsor
Engagement

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning - Sequence
• Sponsor for Performance Improvement
– Clarifying purpose, scope, resources, outcomes
• Conduct the investigation
– Discover the causes, engage with Sponsor, decide when done
• Conduct Learning Session(s) with others
– Provide opportunity to learn and to broaden Insights
• Decide what to address
– Managers select causes to address
• Develop corrective recommendations
– Develop effective solutions to address selected causes
• Action Commitment with Managers
– Gain alignment on recommendations to implement
• Develop and implement solutions
– Managers lead solution implementation
– Check that performance actually improves

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Sponsor Engagement
Benefits

• Keep the Sponsor informed of progress

– What we currently understand
– What we have yet to discover
• Engage the Sponsor in the reasoning / data
• Provide the Sponsor an opportunity to understand
and learn
• Gain agreement on how to proceed (what is next)
– Based on team’s understanding
– Based on Sponsor’s objectives

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Presentation Tips
• General
– Present from posters and charts on the wall
– Select most appropriate presenters for each section
– Start with objectives/agenda
– Practice!
• Incident Description
– Summarize what happened (no causes or recommendations)
• Problem Statement
– Include reasons for selecting that problem statement
– Connect to top box on Cause Effect Diagram
• Timeline
– Explain scale and any changes in scale
– Decide the best sequence to present multiple timelines
– Identify “top box” on timeline

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Presentation Tips
• Cause and Effect Diagram
– Start with top box
– Present one cause effect “level” at a time
– Decide most understandable sequence to present cause boxes for
each level (Tip: number in pencil on chart)
– For small sections of logic, complete them and go back up
– Present what’s written in the boxes (Don’t read data quality)
– State the logic (AND, OR) when reading a group of causes
– If you developed a graphic to explain the “how” of a causal theory,
use it, but always go back to cause effect diagram
– Summarize theories after you present them
– At end summarize overall chart or walk back up chart
– Highlight any questions that are still open
– When asked a question, don’t “make up” an answer (It’s OK to say
that you don’t know)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Presentation Activity
• Meet in your assigned teams
• Review/adjust your team’s work as you see fit
– Prepare to present
• Objectives & agenda for the meeting
• Incident description
• Problem Statement
• System drawing
• Timeline (first team only)
• Cause and Effect Diagram
• Proposal what else to investigate (no
solutions)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Developing Recommendations

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning - Sequence
• Sponsor for Performance Improvement
– Clarifying purpose, scope, resources, outcomes
• Conduct the investigation
– Discover the causes, engage with Sponsor, decide when done
• Conduct Learning Session(s) with others
– Provide opportunity to learn and to broaden Insights
• Decide what to address
– Managers select causes to address
• Develop corrective recommendations
– Develop effective solutions to address selected causes
• Action Commitment with Managers
– Gain alignment on recommendations to implement
• Develop and implement solutions
– Managers lead solution implementation
– Check that performance actually improves

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Developing Recommendations
“Corrective” solution ideas create new desired
effects instead of unwanted effects by
• Addressing existing causes in the system
• Creating new causes in the system

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Development
Effective Solutions
– Are based on new understanding of causes
(what we are learning)
– Address selected causes (on the diagram)
– Create desired outcomes instead
– Work, and are sustainable in current reality
– May require multiple actions in combination
– Usually emerge from several iterations

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Development
Process
1. Decide what cause to address
– What you want to change about the existing system
2. Describe what you want to create instead
– Achieve, avoid, maintain
3. Develop Solution Ideas
4. Test Solution Ideas for Effectiveness
– If the solution had already been in place, how would it
have prevented the incident and what would have been
created instead?
– How effective will the solution be to prevent similar
incidents in the future? Could the solution create
unintended consequences?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Steps 1-2
2 – What you
want to create
instead

Achieve, avoid, maintain

1 - Cause
to address

What you want to change

about the existing system

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
Recommendation Tips (Steps 1-2)
• Be specific
– Which cause on the diagram are you trying to address?
– What exactly are you trying to create?
• Choose a cause away from the top or bottom of
diagram

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Activity (Step 1-2)

As a team, review your cause and effect diagram

and reflect on the OIC’s discussion about the
“system” of the Sara Platform

1. Decide what cause to address (and Why)

– What you want to change about the existing system
2. Describe what you want to create instead
– Achieve, Avoid, Maintain

Record on flipcharts.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Development
Process
1. Decide what cause to address
– What you want to change about the existing system
2. Describe what you want to create instead
– Achieve, avoid, maintain
3. Develop Solution Ideas
4. Test Solution Ideas for Effectiveness
– If the solution had already been in place, how would it
have prevented the incident and what would have been
created instead?
– How effective will the solution be to prevent similar
incidents in the future? Could the solution create
unintended consequences?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Develop Solution Ideas (Step 3)
• Go back to the incident and ask
– “What could have been done differently to address
the selected cause and to create what you want
instead”?

• List solution ideas and any

concerns you may have with
Solution Concerns
Ideas

each idea

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Develop Solution Ideas (Step 3)

Tips:

• Be creative, build on the ideas of others

• Be practical, work in the system
• Avoid adding more work, making more rules.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Activity (Step 3)
• As a team, go back to the incident timeline and cause
tree.
• Develop solution ideas that would address the cause
you selected and create what you want instead.
• List solution ideas and concerns on a flipchart.
• Select a solution idea that seems practical for the Sara
Platform incident.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Development
Process
1. Decide what cause to address
– What you want to change about the existing system
2. Describe what you want to create instead
– Achieve, avoid, maintain
3. Develop Solution Ideas
4. Test Solution Ideas for Effectiveness
A. If the solution had already been in place, how would it
have prevented the incident and what would have been
created instead?
B. How effective will the solution be to prevent similar
incidents in the future? Could the solution create
unintended consequences?

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Testing for Effectiveness (Step 4)
Benefits

Often, our initial solution ideas are

– Based on our normal (old) way of thinking about “how
things should be”
– Not fully effective in creating the desired outcomes

By testing our solution ideas for effectiveness, we can

– Avoid wasting resources on solutions that don’t work
– Increase certainty that solutions will create what we
want

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Activity (Step 4)
As a team, test your solution idea for effectiveness:

4A. If the solution had already been in place, how would

it have prevented the incident and what would have
been created instead?
– Describe when exactly the solution would have interrupted the
incident timeline
– Describe what would most likely have happened instead from
that point forward
– Assess whether the “alternate” timeline would likely create a
desirable outcome

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Recommendation Activity (Step 4)

4B. How effective will the solution be to prevent similar

incidents in the future? Could the solution create
unintended consequences?
– Describe what would have to be true for the solution to prevent
a similar incident in the future
– Assess how likely it is that the solution will be sustainable in
the current system
– Imagine potential unintended consequences created by the
solution
Prepare to report out all of your recommendation
development work.

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving
Performance
from
Causal
Learning

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning
Causal Learning application
• Incident Investigation process
– Sponsored Causal Learning investigation teams
– Investigate one incident at a time
– Identify system cause patterns across incidents
– Pool of skilled Causal Learning Facilitators
• Problem Solving
– Full time problem solving teams or discreet problem
solving efforts
– Causal Learning skills in functions (e.g. Reliability,
Technical Disciplines, Production)

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning
Effectiveness in
• The Causal Learning Method
– Investigation – Finding Cause
– Learning from Cause – Making Meaning
– Developing Corrective Action from Cause
– Implementing to create the performance wanted
instead

• Leadership (Sponsorship)
– Creating Conditions for Performance Improvement
– Leading from Role

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning

Time for the Organization to Shift

• To Reason Causally
• To Learn how to Learn
• To See the current System
• To Create Action that is truly Corrective
• To Implement, creating the Performance
wanted instead

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning - Sequence
• Sponsor for Performance Improvement
– Clarifying purpose, scope, resources, outcomes
• Conduct the investigation
– Discover the causes, engage with Sponsor, decide when done
• Conduct Learning Session(s) with others
– Provide opportunity to learn and to broaden Insights
• Decide what to address
– Managers select causes to address
• Develop corrective recommendations
– Develop effective solutions to address selected causes
• Action Commitment with Managers
– Gain alignment on recommendations to implement
• Develop and implement solutions
– Managers lead solution implementation
– Check that performance actually improves

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.
 Improving Performance from
Causal Learning - Sponsorship
• Sponsoring – creating conditions to ensure performance
improvement outcomes
• Guiding the Reasoning – keeping it causal, specific,
evidence based
• Directing Deeper Level Analysis – understanding system
from causes of human behavior
• Leading Learning – leading the shifting of beliefs
• Deciding what must be addressed – determining the critical
few causes to correct
• Ensuring Recommendations are Corrective – providing
performance correction potential
• Implementing agreed Solutions – gaining return

Copyright, all rights reserved,

Causal Learning
2020 Shell Exploration & Production Co.