The risk know-how framework sets out what is needed to make sense of risk. It has been developed through discussions and interactions with communities and risk experts to facilitate those decisions.

Risk know-how in a community means we can:

  • Ask questions about specific risks
  • Find suitable and reliable risk information
  • Understand how the framing of the information can be manipulated
  • Understand information about the magnitude of a risk and the effectiveness of a response
  • Not be surprised by the consequences of a decision made regarding risks  
  • Make reasonable comparisons between the potential benefits and harms of acting or not acting – understanding the trade-offs
  • Appreciate the danger of following information that confirms what we prefer or believe without challenging it 
  • Be aware that the available risk information can change, and therefore decisions based on it can change too
  • Respect that every person has unique risk and benefit trade-offs, and not everyone has the opportunity to act upon risks
From Risk know-how practitioners

“I want to encourage information literacy – to interrogate assumptions about what is reliable. We start with sources from local, state or regional governments, then go to national governments and international agencies, […] after that we explore the grey literature. I also recommend tertiary information, […] statistics databases that provide and aggregate raw data.”

Chandler Christoffel, User Experience Librarian, USA 

“Sometimes you face the same conditions but make different decisions because of how young the group is, [and] how large the group is. What feels like the right decision one day might feel too risky a different day.”

Sarah Whitaker, Forest Kindergarten Director, USA 

“Fishermen will go out to sea even with a cyclone or a storm approaching because the risk of not catching fish and being able to eat feels more present. “

Sazedul Hoque, Researcher in Food Safety and Fisheries Technology, Bangladesh  

To make sense of risks in real issues we need to:

Clarify the scope of the risk

Outcomes and consequences

Risk of what?

Infection, hospitalization, dying?

The expression “risk of Covid” that was widely used is ambiguous.  

“Risk of Covid” could have different meanings: 

  • Risk of infection: the likelihood of contracting the virus in the first place. 
  • Risk of hospitalization (if already infected): the chance of one’s condition escalating to the point of requiring medical intervention in a hospital.
  • Risk of dying if infected: the chance of dying due to the virus among those infected.
  • Risk of dying if not infected: the chance of dying due to the virus among the population. 

Answering the question ‘risk of what’ helps understand the outcome or consequence of concern.

Show layer:

Risk factors and hazards

What may be contributing to the risk?

Overfishing or a sudden sea temperature spike?

A coastal community relies on fishing as a primary source of food and income. When faced with a reduction in fish population, people blame different things.

To understand the risk we need to know what factors may be contributing to that risk, e.g. are fish numbers down due to overfishing or due to higher sea temperatures?

  • Sudden sea temperature spikes: warmer waters can alter the habitats of many fish species, affecting their breeding patterns and food sources.
  • Overfishing: fishing intensively could mean that fishes are caught faster than they can reproduce.

While problems may have a single cause, more often they result from multiple factors. Identifying all these factors is essential for understanding the complexity of the problem.

Show layer:

Population

Who or what could be affected by that risk?

All children or under 5-year-olds?

Meteorologists have announced the possibility of heavy rainfalls leading to a river overflow.

Some people feel unaffected by the situation as they live ‘far enough’ from the river, convinced that their houses will remain untouched. On the other hand, others believe the entire town might be damaged and want to take preventive measures.

To be able to take action and mitigate the risk we need to know who or what could be affected. In this example it could be:

  • Riverside community: those living close to the river face the most immediate risk of flooding.
  • Whole town: more extensive overflow could even influence residents living farther away, disrupting daily routines and transportation.
  • Infrastructure: essential structures, including bridges and roads, particularly in low-lying zones, might be compromised.
  • Emergency services: their efficiency in addressing calls could be hampered by a sudden surge in aid requests.

Knowing who or what will actually be affected helps decide how to take action or who needs support.

Show layer:

Period of time and cumulative risks

What’s the period of time of the estimated risk?

One day, one year, a lifetime?

In fishing is one of the most dangerous occupations in the world: it was estimated that the risk of dying from accidents is around 132 deaths per 100,000 workers. But is that number per year? Per decade? Per a life time?  

But some fishermen might have a higher risk than others since they are exposed to more risk factors.  

The risk of having an accident in rough seas is higher for a vessel that lacks maintenance, is overloaded and whose workers lack training. The risk of an accident accumulates due to the effect of multiple risk factors. 

Risks are calculated over given timeframes. Risks can accumulate if there are multiple risk factors occurring at the same, or if one of them repeats over time.  

Show layer:

Understand the numbers

Absolute and relative risks

Differentiate between absolute and relative risks

“Contraceptive pills ‘double the risk’ of blood clots.” But what was the risk to start off?

 

 

In 1995, headlines scared thousands of women around the world when they said that contraceptive pills double the risk of blood clots. But ‘double’ is a relative risk. We need to ask “double what?’”

 To understand this: 

Without the pill: Out of 7,000 women, 1 is expected to develop blood clots in a year. (This is the absolute risk in the non-exposed group or baseline risk). 

With the pill: Out of 7,000 women, 2 will likely develop blood clots in a year. This is one additional case per 7,000 users. (This is the absolute risk in the exposed group). 

Cases went from 1 to 2 out of 7000 women so indeed doubled but the absolute risk remains small. 

Relative risks are not enough to guide decisions since  they don’t give any information about the size of the risk – only indicate how larger or smaller is one risk relative to another. For informed decisions, we need to know the absolute risks (probabilities) in all groups compared, e.g. people taking vs not-taking a medicine. 

Show layer:

Expected frequencies and Percentages

Realise that the way probability is expressed can impact our perception of risk

Would you be more inclined to use a dating app if you heard 1 out of 5 couples succeed, or another where 20% does? Same odds, different feel!

The way probabilities are expressed can impact perception. There are three main ways:

  1. Expected frequencies with same denominator (“x in 100” or “x in 1000” or another denominator):
  • e.g., 3 out of 50 vs. 4 out of 50.
  • This format is recommended for clear risk communication
  1. Frequencies with different denominator (“1 in x)”:
  • e.g., 1 in 12 vs. 1 in 17
  • Make people feel the event is a lot more likely to happen
  • Can be challenging for making comparisons.
  • Should be avoided
  1. Percentage:
  • e.g., 6% vs. 8%.
  • Might make people feel the event is less likely to happen
  • Is more difficult to understand when the percentage is lower than 1%. If you use it, add expected frequencies with same denominator (e.g., if you have 0.3% vs 1.2%, add that means 3 out of 1000 vs 12 out of 1000).

Remember! The presentation of numbers can influence risk perception. Generally, it is best to use expected frequencies with same denominator (e.g 1 out of 1000, instead of 0.1%).

Show layer:

Denominator neglect

Don’t just count the incidents

“Titan submarine is safer than flying in a helicopter,” boasted the submarine’s creator short before it tragically imploded. But did he consider how many people choose each adventure?

In 2023, the submarine ‘Titan’ imploded in the ocean killing everyone on board.

The CEO of the company, who lost his life in the incident, had claimed that the Titan was “way safer than flying in a helicopter or even scuba diving” because “there hasn’t even been an injury in 35 years in a non-military submarine.”

Yet, he missed a crucial point: a significantly larger number of people use helicopters and partake in scuba diving every day. He focused solely on the numerators while neglecting the denominators.

The number of incidents may not be enough information, we need to know how many people are exposed to each situation.

Show layer:

Very big and very small numbers

Interpret very big and very small numbers

0.1% might feel too little, 8 million might feel too big, and still refer to the same thing

Think a genetic disease affecting 0.1% of the population is too small to care about? With a world population of roughly 8 billion, that means 8 million people affected.

Conversely, while 8 million people worldwide might suffer from a certain condition this does not mean it is widespread among those you know. It represents 1 in every 1,000 individuals.

Show layer:

Comparisons and context

Be aware of comparisons and context

A construction company is the safest in the area but could still be very dangerous by global standards – don’t be mislead by comparisons!

Construction company A claims to be the safest in its local area– but on a broader scale? 

Out of its 1,000 workers, Company A reports 100 falls every year. In contrast, other companies in the same area report 200 falls for the same number of workers. 

While Company ‘A’ has fewer accidents in this local comparison, 10% of its workforce still experiences accidents annually. When we expand the context from local to national or international safety standards, this rate may stand out as alarmingly high and deemed unacceptable. 

While comparisons can provide some insights, they can also be misleading when using the wrong comparisons.

Show layer:

Averages and Recurrent intervals

Remember what averages can ‘hide’

If half a country faces droughts and the other half floods on average rainfall is ‘as expected’

Country ‘A’ has an average annual rainfall of 20 inches (50cm). This might seem normal for its climatic conditions. However, this average hides that some states barely get any rain, while others get much more than the national average.

Similarly, think about a region that expects a flood once every 10 years – this is an average of the time they are expected, also called recurrent interval. This doesn’t mean that the region will face a flood, then have 9 dry years, and then another flood. It’s possible for the region to have floods two years in a row and then not have another for almost 20 years.

Remember!  Averages and recurrent intervals might mask the true variations and extremes we might encounter. Knowing the type of average and how it was calculated can help understand the fuller picture.

Show layer:

Single event probabilities

Grasp what ‘chance’ means for one-time events

A 30% chance of rain doesn’t mean 30% of the day is wet

Imagine there is a 30% chance of rain where you live.

That means that in 3 out of 10 similar weather scenarios, it rained. It doesn’t mean 30% of the area gets wet or that umbrellas are needed for 30% of the day.

Similarly, when forest officials warn of a 15% chance of wildfires during a particularly dry season, it indicates that in past conditions similar to this, 15 out of 100 times a wildfire sparked. It doesn’t mean that only 15% of the forest will burn.

A one-time event’s chance tells us whether it might occur now; it doesn’t provide information about its duration, magnitude, or past occurrences.

Show layer:

Conditional probabilities

Recognise when one number needs another: conditional probabilities

You receive a bomb threat alert at 2:30am. Given the alert, and considering past system glitches, what’s the real probability it’s an actual threat?

Historically, the system has a false alarm once every 100 alerts, meaning 99 out of 100 times, the alert is real.

However, based on data from the local police, you know that only 1 in 10 real alerts corresponds to an actual threat.

Given the alert and considering past false alarms, what’s the real probability it’s an actual credible threat? The answer depends on two probabilities.

1- Probability that the alert is real given the system’s false alarm rate: 99/100 or 0.99.

2- Probability that a real alert corresponds to a credible threat based on past patterns: 1/10 or 0.10.

Now, the term ‘conditional’ here means that one probability (the threat being credible) is dependent on another condition (the alert being real). Thus, to find out the probability of an actual threat given the alert:

The probability that it’s a real credible threat given an alert is: 0.99 (probability it’s a real alert) x 0.10 (probability a real alert is a credible threat) = 0.099 or 9.9%.

A real alert doesn’t always mean a real threat. The credibility of the threat is ‘conditional’ on the alert being genuine.

Remember! Some probabilities result from the calculation of two probabilities. Understanding this is essential for interpreting certain safety systems and test results.

Show layer:

Think beyond the numbers

Correlation vs. causation

Is it a correlation or causation?

More car accidents occur during hot days but does the heat directly cause the accidents?

Car insurance companies have noticed that there are more car accidents during hot days. While the accidents and the weather might seem connected,  car failure because of the heat is not the main culprit. Instead, the heat increases sleepiness and driving errors, which is the reason behind the accidents.  

Just because two things happen together doesn’t mean one causes the other. 

Show layer:

Adjusting

If a risk was ‘adjusted’, what does that mean?

At first glance, it appeared that social media could enhance happiness. But after adjusting for exercise and other variables, it’s actually the opposite effect.

Researchers found social media use is linked to depression after ‘adjusting’ for age, family income, exercise and drug use. This means they’ve tried to remove the influence of these other factors to focus solely on the impact of social media. Adjusting is helpful, but it’s not definitive proof of causation, as there are always things we might not know or haven’t measured. 

 

How can we know the differences in happiness are actually linked to social media use and not to money, exercise, diet and many other variables? To explore that question, researchers adjust for multiple factors.

Researchers found social media use is linked to depression after ‘adjusting’ for age, family income, exercise and drug use. This means they’ve tried to remove the influence of these other factors to focus solely on the impact of social media. Adjusting is helpful, but it’s not definitive proof of causation, as there are always things we might not know or haven’t measured.

Remember! ‘Adjusting’ in research aims to clarify the relationship between variables, but it doesn’t guarantee a perfect understanding or prove causation. Always be open to unseen factors.

Show layer:

Thresholds

What does ‘safe’ or ‘unsafe’ mean?

Words mean different things to different people

A river in a community faced lead contamination. Authorities declared the water ‘unsafe’ for consumption. Later they announce it’s now ‘safe.’ But what defines safe or unsafe?

In the context of lead in water,  ‘safe’ means zero because lead can be harmful to human health at all exposure levels. ‘Unsafe’ is anything above zero.  

However, for other contaminants such as chlorine, ‘safe’ doesn’t mean zero but a concentration below 4 milligrams per litre.

The same word can have a different meanings.  Numbers can help explain what words actually mean. 

Show layer:

Uncertainty

Consider uncertainty, don’t just avoid it

Should mariners cancel their routes after a typhoon warning? Making a decision requires addressing all uncertainties involved.

Predictions say the typhoon will have winds of 100 km/h. But there are multiple uncertainties: 

  •  About the accuracy of numbers: There’s a margin: it could vary by ±20 km/h.
  • About the quality of the information:Different weather agencies give different forecasts. One sees the typhoon taking a northern route, another says south. It’s important to weigh the significance of these outcomes differently, given that one agency’s estimation relies on a one-year dataset, while the other draws from a 25-year dataset.
  • About the unknowns: There could also sudden changes in direction or strength due to factors not yet in the models.  

 

Show layer:

Balanced information, False balance and Trade-offs

Have all potential benefits and harms been taken into account?

There are plans to build a nuclear energy plan in a community. Some love the idea of promoting a cleaner atmosphere , but others are concerned about accidents and waste.

The regional government and utility companies want to build a nuclear power plant. Some community members think  it will reduce pollution and provide energy. Others worry about accidents and the nuclear waste management 

Whether it’s building a nuclear plant or getting a vaccine, we need a full discussion of potential benefits and harms, and costs, so people can weigh up the trade-offs. 

Show layer:

RISK TO INDIVIDUALS AND RISK TO POPULATIONS

What’s the difference between risk to individuals and risk to populations?

There are 100 car accidents per day in route Z. But how should a daily commuter make sense of this?

Every day, 100 accidents occur on Route Z. This is informative for local authorities who must redesign the route and work on traffic regulations.  

But what about the regular commuter? For an individual contemplating their daily drive on Route Z, a more meaningful measure for personal decision-making is this: 1 in 5,000 commuters on Route Z experience an accident each day. 

Of course, individual risk varies and it’s not possible to predict the risk of a specific person. For instance, if you’re not adhering to traffic regulations or if you’re a less attentive driver, your personal risk might be higher. But understanding both the population-wide impact and the individual’s perspective gives a more holistic view of the situation. 

A risk can be very concerning from a population perspective yet seem minimal from an individual perspective, and vice versa. 

Show layer:

Cognitive bias and Patterns in randomness

You can’t rely on your instincts

Just because it hasn’t happened, it doesn’t mean it won’t. And just because you see a pattern, it doesn’t mean it’s real.

In a Mexico City neighbourhood, Sr. García dismissed advice to inspect his home despite visible cracks, confidently stating, “My house has stood for decades.” A year later, an unexpected earthquake struck, impacting his neighbourhood, including his house.

Risks don’t necessarily adhere to our perceptions or past experiences. Just because you’ve been safe in the past doesn’t mean you always will be.

Show layer:

Quality of evidence

How good is the quality of the evidence and can I rely on it?

A resort over mangroves, backed by a study claiming no harm to wildlife. But who conducted the study and how?

Politicians decided to support the destruction of mangroves to build up a resort hotel claiming a study found wild life wouldn’t be affected. When looking at the study:

  • they only observed a small fraction of the area,
  • they collected data during the migrating season of many animal species,
  • the study was conducted by a private group of researchers linked to the hotel company.

The quality of evidence depends on the source, methods, potential biases of a study, and other aspects that should be considered before incorporating the information into decision-making.

Show layer: