- Press Release Distribution

Want better AI? Get input from a real (human) expert

A dash of human expertise yields smarter machine learning, higher confidence scores

Want better AI? Get input from a real (human) expert
( RICHLAND, Wash.—Can AI be trusted? The question pops up wherever AI is used or discussed—which, these days, is everywhere.


It’s a question that even some AI systems ask themselves.


Many machine-learning systems create what experts call a “confidence score,” a value that reflects how confident the system is in its decisions. A low score tells the human user that there is some uncertainty about the recommendation; a high score indicates to the human user that the system, at least, is quite sure of its decisions. Savvy humans know to check the confidence score when deciding whether to trust the recommendation of a machine-learning system.


Scientists at the Department of Energy’s Pacific Northwest National Laboratory have put forth a new way to evaluate an AI system’s recommendations. They bring human experts into the loop to view how the ML performed on a set of data.  The expert learns which types of data the machine-learning system typically classifies correctly, and which data types lead to confusion and system errors. Armed with this knowledge, the experts then offer their own confidence score on future system recommendations.


The result of having a human look over the shoulder of the AI system? Humans predicted the AI system’s performance more accurately.


Minimal human effort—just a few hours—evaluating some of the decisions made by the AI program allowed researchers to vastly improve on the AI program’s ability to assess its decisions. In some analyses by the team, the accuracy of the confidence score doubled when a human provided the score.


The PNNL team presented its results at a recent meeting of the Human Factors and Ergonomics Society in Washington, D.C., part of a session on human-AI robot teaming.

“If you didn’t develop the machine-learning algorithm in the first place, then it can seem like a black box,” said Corey Fallon, the lead author of the study and an expert in human-machine interaction. “In some cases, the decisions seem fine. In other cases, you might get a recommendation that is a real head-scratcher. You may not understand why it’s making the decisions it is.”



The grid and AI


It’s a dilemma that power engineers working with the electric grid face. Their decisions based on reams of data that change every instant keep the lights on and the nation running. But power engineers may be reluctant to turn over decision-making authority to machine-learning systems.

“There are hundreds of research papers about the use of machine learning in power systems, but almost none of them are applied in the real world. Many operators simply don’t trust ML. They have domain experience—something that ML can’t learn,” said coauthor Tianzhixi “Tim” Yin.


The researchers at PNNL, which has a world-class team modernizing the grid, took a closer look at one machine-learning algorithm applied to power systems. They trained the SVM (support-vector machine) algorithm on real data from the grid’s Eastern Interconnection in the U.S. The program looked at 124 events, deciding whether a generator was malfunctioning, or whether the data was showing other types of events that are less noteworthy.


The algorithm was 85% reliable in its decisions. Many of its errors occurred when there were complex power bumps or frequency shifts. Confidence scores created with a human in the loop were a marked improvement over the system’s assessment of its own decisions. The human expert’s input predicted the algorithm’s decisions with much greater accuracy.


More human, better machine learning

Fallon and Yin call the new score an “Expert-Derived Confidence” score, or EDC score.


They found that, on average, when humans weighed in on the data, their EDC scores predicted model behavior that the algorithm’s confidence scores couldn’t predict.


“The human expert fills in gaps in the ML’s knowledge,” said Yin. “The human provides information that the ML did not have, and we show that that information is significant. The bottom line is that we’ve shown that if you add human expertise to the ML results, you get much better confidence.”

The work by Fallon and Yin was funded by PNNL through an initiative known as MARS—Mathematics for Artificial Reasoning in Science. The effort is part of a broader effort in artificial intelligence at PNNL. The initiative brought together Fallon, an expert on human-machine teaming and human factors research, and Yin, a data scientist and an expert on machine learning.


“This is the type of research needed to prepare and equip an AI-ready workforce,” said Fallon. “If people don’t trust the tool, then you’ve wasted your time and money. You’ve got to know what will happen when you take a machine learning model out of the laboratory and put it to work in the real world.


“I’m a big fan of human expertise and of human-machine teaming. Our EDC scores allow the human to better assess the situation and make the ultimate decision.”

# # #


[Attachments] See images for this press release:
Want better AI? Get input from a real (human) expert Want better AI? Get input from a real (human) expert 2


Boomerang-like beams of light

Boomerang-like beams of light
Researchers at the University of Warsaw's Faculty of Physics have superposed two light beams twisted in the clockwise direction to create anti-clockwise twists in the dark regions of the resultant superposition. The results of the research have been published in the prestigious journal “Optica”. This discovery has implications for the study of light-matter interactions and represents a step towards the observation of a peculiar phenomenon known as a quantum backflow. “Imagine that you are throwing a tennis ball. The ball starts moving forward with positive momentum. If the ball doesn’t hit an obstacle, you are unlikely to expect it to suddenly ...

Miniature colons with immune components aid the study of intestinal diseases

Miniature colons with immune components aid the study of intestinal diseases
A team at the Medical University of South Carolina and Cincinnati Children’s has developed a sophisticated model for studying the diseased colon that could lead to the development of personalized treatments for colon-related diseases, such as cancer and inflammatory bowel disease (IBD). The researchers report their findings in the Nov. 2 issue of Cell Stem Cell. MUSC Hollings Cancer Center researcher Jorge Munera, Ph.D., collaborated with James Wells, Ph.D., and Daniel Kechele, Ph.D., both of Cincinnati Children’s, to grow miniature human colons complete with an immune system in the lab. This model improves upon existing organoids, or mini ...

Are vanadium flow batteries worth the hype? (video)

Are vanadium flow batteries worth the hype? (video)
WASHINGTON, Nov. 20, 2023 — There’s a century-old technology that’s taking the grid-scale battery market by storm. Based on water, virtually fireproof, easy to recycle and cheap at scale, vanadium flow batteries could be the wave of the future. Reactions is a video series produced by the American Chemical Society and PBS Digital Studios. Subscribe to Reactions at and follow us on Twitter @ACSReactions. The American Chemical Society (ACS) is a nonprofit organization chartered by the U.S. Congress. ACS’ mission is to advance the broader chemistry enterprise ...

These bats use their penis as an “arm” during sex but not for penetration

These bats use their penis as an “arm” during sex but not for penetration
Mammals usually mate via penetrative sex, but researchers report November 20 in the journal Current Biology that a species of bat, the serotine bat, (Eptesicus serotinus) mates without penetration. This is the first time non-penetrative sex has been documented in a mammal. The bats’ penises are around seven times longer than their partners’ vaginas and have a “heart-shaped” head that is seven times wider than the vaginal opening. Both the penises’ size and shape would make penetration post-erection impossible, and the researchers show that, rather than functioning as a penetrative ...

AI system self-organizes to develop features of brains of complex organisms

Cambridge scientists have shown that placing physical constraints on an artificially-intelligent system – in much the same way that the human brain has to develop and operate within physical and biological constraints – allows it to develop features of the brains of complex organisms in order to solve tasks. As neural systems such as the brain organise themselves and make connections, they have to balance competing demands. For example, energy and resources are needed to grow and sustain the network in physical ...

Half of tested caviar products from Europe are illegal, and some aren’t even caviar

Half of tested caviar products from Europe are illegal, and some aren’t even caviar
Wild caviar, a pricey delicacy made from sturgeon eggs, has been illegal for decades since poaching brought the fish to the brink of extinction. Today, legal, internationally tradeable caviar can only come from farmed sturgeon, and there are strict regulations in place to help protect the species. However, by conducting genetic and isotope analyses on caviar samples from Bulgaria, Romania, Serbia, and Ukraine—nations bordering the remaining wild sturgeon populations—a team of sturgeon experts found evidence that these regulations are actively being broken. Their results, ...

Physicists answer question of Supergalactic Plane’s absent spiral galaxies

Physicists answer question of Supergalactic Plane’s absent spiral galaxies
Astrophysicists say they have found an answer to why spiral galaxies like our own Milky Way are largely missing from a part of our Local Universe called the Supergalactic Plane. The Supergalactic Plane is an enormous, flattened structure extending nearly a billion light years across in which our own Milky Way galaxy is embedded. While the Plane is teeming with bright elliptical galaxies, bright disk galaxies with spiral arms are conspicuously scarce. Now an international team of researchers, co-led by Durham University, UK, and the University of Helsinki, Finland, say different distributions of elliptical and disk ...

Social determinants of health and cardiologist involvement in the care of adults hospitalized for heart failure

About The Study: This study of 1,000 participants found that adults with low household income were less likely than adults with higher incomes to have a cardiologist involved in their care during a hospitalization for heart failure. These findings suggest that socioeconomic status may bias the care provided to patients hospitalized for heart failure.  Authors: Parag Goyal, M.D., M.Sc., of Weill Cornell Medicine in New York, is the corresponding author.  To access the embargoed study: Visit our For The Media website at this link  (doi:10.1001/jamanetworkopen.2023.44070) Editor’s Note: Please see the article ...

Infertility and risk of autism spectrum disorder in children

About The Study: In this study of 1.3 million children from Ontario, Canada, a slightly higher risk of autism spectrum disorder was observed in children born to individuals with infertility, which appears partly mediated by certain obstetrical and neonatal factors. To optimize child neurodevelopment, strategies should further explore these other factors in individuals with infertility, even among those not receiving fertility treatment.  Authors: Maria P. Velez, M.D., Ph.D., of Queen’s University in Kingston, Ontario, Canada, is the corresponding author.  To access the ...

App-based interventions for moderate to severe depression

About The Study: In this systematic review and meta-analysis of 13 randomized clinical trials of app interventions with 1,470 participants, the feasibility and efficacy of mobile app interventions were supported in treating moderate and severe depression, and practical implications were also provided for developing effective app-based interventions in clinical practice.  Authors: Ji-Won Hur, Ph.D., of Korea University in Seoul, Republic of Korea, is the corresponding author.  To access the embargoed study: Visit our For ...


Bioinformatics approach could help optimize soldiers’ training for improved readiness and recovery

Earth scientists describe a new kind of volcanic eruption

Warmer wetter climate predicted to bring societal and ecological impact to the Tibetan Plateau

Feeding infants peanut products protects against allergy into adolescence

Who will like beetle skewers? What Europeans think about alternative protein food

ETRI wins ‘iF Design Award’ for mobile collaborative robot

Combating carbon footprint: novel reactor system converts carbon dioxide into usable fuel

Investigating the origin of circatidal rhythms in freshwater snails

Altering cellular interactions around amyloid plaques may offer novel Alzheimer’s treatment strategies

Brain damage reveals part of the brain necessary for helping others

Surprising properties of elastic turbulence discovered

Study assesses cancer-related care at US hospitals predominantly serving minority populations compared with non-minority serving hospitals

First in-human investigator-initiated clinical trial to launch for refractory prostate cancer patients: Novel alpha therapy targets prostate-specific membrane antigen

Will generative AI change the way universities communicate?

Artificial Intelligence could help cure loneliness, says expert

Echidnapus identified from an ‘Age of Monotremes’

Semaglutide may protect kidney function in individuals with overweight or obesity and cardiovascular disease

New technique detects novel biomarkers for kidney diseases with nephrotic syndrome

Political elites take advantage of anti-partisan protests to disrupt politics

Tiny target discovered on RNA to short-circuit inflammation, UC Santa Cruz researchers find

Charge your laptop in a minute? Supercapacitors can help; new research offers clues

Scientists discover CO2 and CO ices in outskirts of solar system

Theory and experiment combine to shine a new light on proton spin

PKMYT1, a potential ‘Achilles heel’ of treatment resistant ER+ breast cancers with the poorest prognosis

PH-binding motifs as a platform for drug design: Lessons from protease-activated receptors (PARs)

Virginia Tech researcher creates new tool to move tiny bioparticles

On repeat: Biologists observe recurring evolutionary changes, over time, in stick insects

Understanding a broken heart

Genetic cause of rare childhood immune disorders discovered

With wobbling stars, astronomers gauge mass of 126 exoplanets and find 15 new ones

[] Want better AI? Get input from a real (human) expert
A dash of human expertise yields smarter machine learning, higher confidence scores