Human error (Reason, 1990)

Error-free performance

The odds against error-free performance are high.
There is one way of performing a task correctly, but
each step provides a multitude of inappropriate pathways.

Computer error (Ensmenger, 2010)

By the end of 1960s managers, academics and governments were talking of a crisis. In the next several decades, they released warnings about the desperate state of the software industry with ritualistic regularity.

How often do computers kill people?

Mechanizing Proof (MacKenzie, 2001)

Collected around 1,100 deaths

90% caused by faulty human-computer interaction

4% caused by physical faults (interference)

3% caused by software bugs (mostly Patriot missile)

Poor interfaces and organizational failings

Faulty human-computer interaction

Human error usually is a result of poor design. Human err continually; it is an intrinsic part
of our nature. System design should take this
into account.

(Norman, 1988)

How and why of error research

Why we need to understand errors

• Design resilient systems
• Criminal responsibility implications
• Teach us how cognition works

Each step provides a multitude of inappropriate pathways

Many kinds of errors

When error occurs

• Perceptual errors
• Cognitive errors
• Motor errors

How error occurs

• Was the action planned?
• What caused the error?
• Mistakes, failures, slips and lapses

Perceptual, cognitive and motor errors

Perceptual error
Misleading clues or sensory overload
Did not notice the batter indicator

James Reason (1990)

Human Error

Four basic kinds of errors

Mistake - Following a wrong plan
Application of a bad rule or poor remembering

Failure - Good plan, but too hard
Shooting all the monsters in a game

Lapse - Working memory failure
I forgot what I was just doing

Slip - Failed execution
Automatically getting off at your usual station

What kind of error is this?

Mistake, failure, lapse or slip?

Buying a wrong train ticket

Wrong ticket by mistake

• Cannot figure out what to buy
• Have a rail card?
• When is super off peak?

Wrong ticket as a slip

• I always buy one particular ticket
• Doing the same over weekend
• The bloody machine resets your options!

Buying a wrong train ticket

Why are we classifying kinds of errors?

Novices make mistakes

Do not understand how rail tickets work

Experts make slips

Accidentally click on a wrong button

Many kinds of slips and lapses

Empirical clustering of common errors

1. Capture errors
2. Description errors
3. Data driven errors
4. Associate activation
5. Loss of activation
6. Mode errors

Capture errors

Capture error example
I was using a copying machine, and I was counting the pages. I found myself counting, "1, 2, 3, 4, 5, 6, 7, 8, 9, 10, Jack, Queen, King."
I had been playing cards recently.

How capture errors happen?
A more familiar sequence of action captures a less familiar sequence of actions

Description errors

Description error example
Correct action using a
similar but wrong object

How they happen?
Accidentally perform different action with similar description

Data-driven and associate errors

Data driven error

• Sensory information intrudes
• Saying X because someone said it

Associate activation error

• Internal thought intrudes
• Enter login for another system
• Dismiss an important message box

Loss of activation and mode errors

Loss of activation

• Forgot the goal in the middle
• Go to Google but forget what you wanted to search for

Mode error

• Do action in a wrong mode
• Typing WITH CAPS LOCK ON
• Crash an airplane...

Air Inter Flight 148

Crashed at 19:20:33 CET in the mountains at an altitude of 800 m

Controls with two descent modes

1. Angle in degrees
2. Feet per minute

Not the only cause, but one factor..

Angle in degrees

Descent by
3.2 degrees

Feet per minute

Descent by 3200 feet per minute

Detecting errors

How to detect errors

• Self-monitoring
• Environment gives us hints
• Another person notices it

Why we miss errors

• Accept simplistic partial explanation
• Minor mismatch of mental model
• Relevance bias prevents us from seeing big picture

Detecting errors

Why is error detection getting harder

Experienced mechanics notice errors by sound

In complex systems you cannot see errors directly

Latent errors more common than active errors

Swiss cheese model of complex errors

Resident pathogen metaphor for system errors

Cheese model

Many factors contribute to an error

All have to be aligned for accident to happen

System response to errors

How can system respond to a detected error

Do nothing - Leave everything to the user

Gagging - Prevent unrealistic intentions

Warnings - Warn about unrealistic intentions

Self-correct - Automatically do the right thing

Dialogue - Ask user to explain or clarify

Designing for avoiding errors

Human error usually is a result of poor design. Human err continually; it is a part of our nature. System design should take this into account.

Designing for avoiding errors

Consider conversation between two people

Grammatical errors are not even noticed

Restarts and corrections are ignored

Misstatements interpreted approximately

Statements checked for common sense

Confusion corrected via a dailogue

Designing for avoiding errors

What can computer system do?

Understand root causes behind errors

Implement common sense checks for inputs

Allow undoing an action

Support error discovery & allow corrections

Do not reject, but allow to correct

Buying a London Canterbury ticket

Many ticket types

Time and date sometimes matter

Railcard discounts

Buying a London Canterbury ticket

Ticket options at 9:59am

Changing time used to change options

How to improve ticket machines?

Ticket options at 10:00am

Do not add railcard first, it would get removed

How to improve ticket machines?

Errare humanum est

If it is impossible to guarantee the elimination of errors, then we must discover ways of mitigating their consequences in unforgiving situations.

What to do when errors happen

Learning to live with errors

• Users learn to avoid errors
• Do not hide too many details
• Leave room for workarounds

Resilience engineering

• Monitoring so that experts can notice
• Allow manual correction
• Analyze developments in advance

Different kinds of evaluation

Different types of errors
Mistakes, failures, lapses, slips
Experts make slips, novices mistakes

Two ways of grouping errors
Was the action planned? What went wrong?
Analysis of common errors empirically

How to prevent and correct errors
Design tips to prevent errors
Different ways of responding to errors

CO582: Errors

What you should remember from this lecture

• Recognize mistakes, failures, slips and lapses
• Capture, description, data, association & mode errors
• How can systems prevent & respond to errors

Tomas Petricek
t.petricek@kent.ac.uk | @tomaspetricek

Books

• Human Error, James Reason
• The Design of Everyday Things, Don Norman

Papers and links

• Experimental Study of Vertical Flight Path Mode Awareness, Eric N. Johnson, Amy R. Pritchett