​The '85% Rule': Why a dose of failure optimizes learning

If you're always succeeding, you're probably not learning much.

​The '85% Rule': Why a dose of failure optimizes learning
Photo by Minas Panagiotakis/Getty Images
  • A recent study examined the rates at which machine-learning algorithms learned to recognize images of tumors.
  • The results showed that learning was optimized when the algorithms guessed incorrectly about 15 percent of the time.
  • The researchers suggested that their findings apply to human and animal learning, too.


In learning, most people intuitively recognize that a bit of a challenge is a good thing. The task shouldn't be too hard, nor too easy. This conventional wisdom explains, for example, why the levels of a video game become incrementally more difficult, or why a piano instructor would choose to teach a beginning student "Twinkle Twinkle Little Star" instead of a Chopin Étude.

But exactly how difficult should learning be? Is there a "sweet spot"?

The answer seems to be yes, according to a recent study which found learning is optimized when the learner gets it right about 85 percent of the time. To get that number, scientists trained machine-learning algorithms to recognize images of tumors at various levels of difficulty. They found that the algorithms learned most efficiently when the failure rate was about 15 percent.

The 85% rule for machines and humans

"These ideas that were out there in the education field — that there is this 'zone of proximal difficulty,' in which you ought to be maximizing your learning – we've put that on a mathematical footing," Robert Wilson, an assistant professor of psychology and cognitive science at the University of Arizona, and lead author of the study, told UA News. "If you have an error rate of 15% or accuracy of 85%, you are always maximizing your rate of learning in these two-choice tasks."

Of course, the study involved algorithms, not humans. However, the researchers wrote that their findings also describe optimal learning in humans and animals, "from perception, to motor control to reinforcement learning." In the study, the researchers tweaked their model to reflect the ways in which monkeys learn a task over time. The results showed that, in all scenarios, learning was optimized with an accuracy rate of about 85 percent.

Lung cancer, MRI

Photo by: BSIP/Universal Images Group via Getty Images

Wilson said the 85 percent rule would be particularly applicable in perceptual learning, in which we gradually learn tasks by interacting with the environment, such as learning to identify tumors in images.

"You get better at figuring out there's a tumor in an image over time, and you need experience and you need examples to get better," Wilson said. "I can imagine giving easy examples and giving difficult examples and giving intermediate examples. If I give really easy examples, you get 100% right all the time and there's nothing left to learn. If I give really hard examples, you'll be 50% correct and still not learning anything new, whereas if I give you something in between, you can be at this sweet spot where you are getting the most information from each particular example."

Grit and flow states

But there's another reason why it's important for us to incorporate a healthy dose of failure into learning: it prepares people for the inevitable challenges of life. Tom Hoerr, former leader of the New City School in St. Louis, Mo., said students need to learn not only curriculum, but also the emotional tools necessary to withstand challenges.

"If our kids have graduated from here with nothing but success, then we have failed them, because they haven't learned how to respond to frustration and failure,"Hoerr told KQED.

There's also reason to think following the 85 percent rule could help people enter a flow state — a feeling of being "in the zone" that occurs when you're fully immersed in a task that's appropriately challenging.

"Boredom is where you're not learning, and your accuracy is at 100 percent," Wilson told Psychology Today. "And anxiety is where you're not learning, and your accuracy is at 50 percent or chance. This is pure speculation, but that's something we're excited to think about going forward."

U.S. Navy controls inventions that claim to change "fabric of reality"

Inventions with revolutionary potential made by a mysterious aerospace engineer for the U.S. Navy come to light.

U.S. Navy ships

Credit: Getty Images
Surprising Science
  • U.S. Navy holds patents for enigmatic inventions by aerospace engineer Dr. Salvatore Pais.
  • Pais came up with technology that can "engineer" reality, devising an ultrafast craft, a fusion reactor, and more.
  • While mostly theoretical at this point, the inventions could transform energy, space, and military sectors.
Keep reading Show less

Why so gassy? Mysterious methane detected on Saturn’s moon

Scientists do not know what is causing the overabundance of the gas.

An impression of NASA's Cassini spacecraft flying through a water plume on the surface of Saturn's moon Enceladus.

Credit: NASA
Surprising Science
  • A new study looked to understand the source of methane on Saturn's moon Enceladus.
  • The scientists used computer models with data from the Cassini spacecraft.
  • The explanation could lie in alien organisms or non-biological processes.
Keep reading Show less

CRISPR therapy cures first genetic disorder inside the body

It marks a breakthrough in using gene editing to treat diseases.

Credit: National Cancer Institute via Unsplash
Technology & Innovation

This article was originally published by our sister site, Freethink.

For the first time, researchers appear to have effectively treated a genetic disorder by directly injecting a CRISPR therapy into patients' bloodstreams — overcoming one of the biggest hurdles to curing diseases with the gene editing technology.

The therapy appears to be astonishingly effective, editing nearly every cell in the liver to stop a disease-causing mutation.

The challenge: CRISPR gives us the ability to correct genetic mutations, and given that such mutations are responsible for more than 6,000 human diseases, the tech has the potential to dramatically improve human health.

One way to use CRISPR to treat diseases is to remove affected cells from a patient, edit out the mutation in the lab, and place the cells back in the body to replicate — that's how one team functionally cured people with the blood disorder sickle cell anemia, editing and then infusing bone marrow cells.

Bone marrow is a special case, though, and many mutations cause disease in organs that are harder to fix.

Another option is to insert the CRISPR system itself into the body so that it can make edits directly in the affected organs (that's only been attempted once, in an ongoing study in which people had a CRISPR therapy injected into their eyes to treat a rare vision disorder).

Injecting a CRISPR therapy right into the bloodstream has been a problem, though, because the therapy has to find the right cells to edit. An inherited mutation will be in the DNA of every cell of your body, but if it only causes disease in the liver, you don't want your therapy being used up in the pancreas or kidneys.

A new CRISPR therapy: Now, researchers from Intellia Therapeutics and Regeneron Pharmaceuticals have demonstrated for the first time that a CRISPR therapy delivered into the bloodstream can travel to desired tissues to make edits.

We can overcome one of the biggest challenges with applying CRISPR clinically.

—JENNIFER DOUDNA

"This is a major milestone for patients," Jennifer Doudna, co-developer of CRISPR, who wasn't involved in the trial, told NPR.

"While these are early data, they show us that we can overcome one of the biggest challenges with applying CRISPR clinically so far, which is being able to deliver it systemically and get it to the right place," she continued.

What they did: During a phase 1 clinical trial, Intellia researchers injected a CRISPR therapy dubbed NTLA-2001 into the bloodstreams of six people with a rare, potentially fatal genetic disorder called transthyretin amyloidosis.

The livers of people with transthyretin amyloidosis produce a destructive protein, and the CRISPR therapy was designed to target the gene that makes the protein and halt its production. After just one injection of NTLA-2001, the three patients given a higher dose saw their levels of the protein drop by 80% to 96%.

A better option: The CRISPR therapy produced only mild adverse effects and did lower the protein levels, but we don't know yet if the effect will be permanent. It'll also be a few months before we know if the therapy can alleviate the symptoms of transthyretin amyloidosis.

This is a wonderful day for the future of gene-editing as a medicine.

—FYODOR URNOV

If everything goes as hoped, though, NTLA-2001 could one day offer a better treatment option for transthyretin amyloidosis than a currently approved medication, patisiran, which only reduces toxic protein levels by 81% and must be injected regularly.

Looking ahead: Even more exciting than NTLA-2001's potential impact on transthyretin amyloidosis, though, is the knowledge that we may be able to use CRISPR injections to treat other genetic disorders that are difficult to target directly, such as heart or brain diseases.

"This is a wonderful day for the future of gene-editing as a medicine," Fyodor Urnov, a UC Berkeley professor of genetics, who wasn't involved in the trial, told NPR. "We as a species are watching this remarkable new show called: our gene-edited future."

Quantcast