About the 'Risk of automation' scores
What is this score?
This score is based on the abilities, knowledge, skills and activities required in order to do the job. The higher it is, the more at risk this occupation is.
It's not a perfect measure by any means, but it can be useful for comparing different occupations. For example, an occupation with a score of 10% is very likely to be at much less at risk compared to an occupation with a score of 90%
How do we calculate this score?
First, we hand-label occupations we are fairly sure are either very easy to automate, or extremely difficult. We also make use of our vast array of user votes to verify some of our assumptions.
We then input a vast array of different attributes for each occupation into a Machine Learning system, which trains a model for us.
We can then use this model to predict how susceptible any occupation is to automation.
We've tested using many different models, and our final results use a Sdca Regression Trainer, which gives us an R-Squared of 0.9193
Many checks have been performed to validate that our model is functioning as reliably as possible, and it's something we are constantly working to improve upon.
What occupations have we hand labelled?
Some examples of occupations we have labelled as very likely to be automated are:
- Data Entry Keyers
- Taxi Drivers
Some examples of those we've labelled as very difficult to automate are:
- Chief Executives
What attributes do we use?
We use data from 0*NET. Some examples of attributes we feed into the model include:
- Thinking Creatively
- Social Perceptiveness
- Assisting and Caring for Others
O*NET gives us 2 values for each attribute: importance, and level.
The "level" rating helps us understand the capabilities required of computer-controlled equipment to perform the tasks of a specific occupation.
For example, low "level" rating for "Manual Dexterity" might correspond to a task like screwing a light bulb into a light socket, while a high "level" rating for the same attribute might be required for performing open-heart surgery with surgical instruments. The specific examples provided for each level give us an idea of the level of manual dexterity needed for an occupation.
What makes an occupation less likely to be at risk of automation?
There are several characteristics that may make an occupation less likely to be at risk of automation:
- Jobs that require a high degree of creativity, judgment, and decision-making are typically more difficult to automate. For example, jobs such as architects, mechanical engineers, and lawyers are less likely to be automated.
- Jobs that involve a high level of personal interaction and emotional intelligence are also less likely to be automated. Examples include jobs such as social workers, and mental health counselors.
- Jobs that involve a high level of physical dexterity and manual skill, such as surgeons, are also less likely to be automated.
- Jobs that require a deep understanding of context and cultural knowledge, such as anthropologists and historians, are also less likely to be automated.
- Jobs that involve working in complex, unstructured environments, are also less likely to be automated, such as forestry workers.
Our model shows that these are the top 10 attributes most correlated with occupations that are difficult to automate:
- Fluency of Ideas
- Social Perceptiveness
- Learning Strategies
- Scheduling Work and Activities
- Thinking Creatively
- Active Learning
- Complex Problem Solving
The occupation I am interested in isn't listed, can you produce a calculation for it?
Unfortunately, we are unable to. If there is no score for it, it's most likely because O*NET don't collect data for it. It may be worth checking again in the future, as their database is constantly being fine-tuned, and we update ours as and when they do.
My occupation gets a really high risk score, should I be worried?
Every situation is unique, there is no one size fits all answer!
The thing we would always advise is to do everything in your control to make sure you stay ahead. Here are a few pointers...
- Stay up-to-date on developments in automation and technology: By staying informed, you can get a sense of which industries and jobs are most likely to be impacted by automation, and take steps to adapt accordingly.
- Keep your skills and knowledge current: By continually learning and acquiring new skills, you can increase your value as an employee and make yourself less vulnerable to automation.
- Consider a career in a field that is less likely to be automated. Changing careers can be a great opportunity to pursue your passions, take on new challenges, learn new skills, increase your job satisfaction and earning potential, and make a positive impact on the world.
- Consider starting your own business: Entrepreneurship can provide a sense of control and stability in an uncertain job market.
- Seek out job opportunities with companies that are committed to responsible automation: Some companies are taking steps to ensure that their adoption of automation is done in a way that is sensitive to the needs and concerns of their employees.
Also, keep in mind that automation can have both positive and negative impacts on employment, and the overall impact is likely to be complex.
Why have some ratings changed over time?
The website originally showed predictions taken from a 2013 study, but the current one's have been calculated in-house.
A multitude of occupations now display significantly different scores. These variations can primarily be traced back to two fundamental causes.
Firstly, the information pertaining to the activities of the occupations has undergone substantial changes since 2013. This is largely due to O*NET's constant data refinement processes. Our methodology, fueled by the significant advancements in CPU power over the past decade, incorporates a larger data set than was previously used.
Secondly, the classifications of certain occupations were reevaluated during our computational process. We chose to exclude some occupations that the Oxford Martin team had labeled as possessing a complete certainty or complete improbability of automation. Simultaneously, we decided to include some occupations previously disregarded by Oxford Martin in our predictive modeling.
The feedback we've collected, with over 100,000 votes to date, also played a significant role in shaping our predictions. This is a resource that the Oxford Martin team did not have available, as they solely relied on the expertise and opinions of their team members.