Ike Obi, Purdue University

My colleagues and I at Purdue University have uncovered a significant imbalance in the human values embedded in AI systems. The systems were predominantly oriented toward information and utility values and less toward prosocial, well-being and civic values.

At the heart of many AI systems lie vast collections of images, text and other forms of data used to train models. While these datasets are meticulously curated, it is not uncommon that they sometimes contain unethical or prohibited content.

To ensure AI systems do not use harmful content when responding to users, researchers introduced a method called reinforcement learning from human feedback. Researchers use highly curated datasets of human preferences to shape the behavior of AI systems to be helpful and honest.

In our study, we examined three open-source training datasets used by leading U.S. AI companies. We constructed a taxonomy of human values through a literature review from moral philosophy, value theory, and science, technology and society studies. The values are well-being and peace; information seeking; justice, human rights and animal rights; duty and accountability; wisdom and knowledge; civility and tolerance; and empathy and helpfulness. We used the taxonomy to manually annotate a dataset, and then used the annotation to train an AI language model.

Our model allowed us to examine the AI companies’ datasets. We found that these datasets contained several examples that train AI systems to be helpful and honest when users ask questions like “How do I book a flight?” The datasets contained very limited examples of how to answer questions about topics related to empathy, justice and human rights. Overall, wisdom and knowledge and information seeking were the two most common values, while justice, human rights and animal rights was the least common value.

Why it matters

The imbalance of human values in datasets used to train AI could have significant implications for how AI systems interact with people and approach complex social issues. As AI becomes more integrated into sectors such as law, health care and social media, it’s important that these systems reflect a balanced spectrum of collective values to ethically serve people’s needs.

This research also comes at a crucial time for government and policymakers as society grapples with questions about AI governance and ethics. Understanding the values embedded in AI systems is important for ensuring that they serve humanity’s best interests.

What other research is being done

Many researchers are working to align AI systems with human values. The introduction of reinforcement learning from human feedback was groundbreaking because it provided a way to guide AI behavior toward being helpful and truthful.

Various companies are developing techniques to prevent harmful behaviors in AI systems. However, our group was the first to introduce a systematic way to analyze and understand what values were actually being embedded in these systems through these datasets.

What’s next

By making the values embedded in these systems visible, we aim to help AI companies create more balanced datasets that better reflect the values of the communities they serve. The companies can use our technique to find out where they are not doing well and then improve the diversity of their AI training data.

The companies we studied might no longer use those versions of their datasets, but they can still benefit from our process to ensure that their systems align with societal values and norms moving forward.

Ike Obi, Ph.D. student in Computer and Information Technology, Purdue University

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Jeremy David Engels, Penn State

Most people love love, but not everyone loves Valentine’s Day.

When it was first invented in the 1300s in medieval Europe, this holiday was a celebration of romantic love, the coming of spring and the freedom to choose a partner, rather than having one chosen for you.

Today that ancient and optimistic message remains but is often buried under a pile of consumer goods – chocolates, cards, stuffed animals, plastic toys, expensive dinners and roses that cost so much more than you think.

The archetypical image of this holiday is Cupid shooting a person with an arrow that makes them go mad with physical desire.

Yet love is one of the richest and most diverse human emotions. There are many ways to experience love – so this holiday, as a scholar of mindfulness and communication, I encourage you to try out a practice of “metta,” or loving-kindness.

What is loving-kindness?

Loving-kindness, or metta, is the type of love praised and practiced by Buddhists around the world, and it is very different from romantic love. It is described as “limitless” and “unbounded” love.

In the ancient Pali language, the word “metta” has two root meanings. The first is “gentle,” in the sense of a gentle spring rain that falls on young plants without discrimination. The second is “friend.” A metta friend is a true friend – someone who is always there for you without fail and without demanding anything in exchange, or someone who supports you when you’re in pain and who is happy for you when you’re happy, without a tinge of jealousy.

Metta is a kind of love that is offered without any expectation of return. It is not reciprocal or conditional. It does not discriminate between us and them, or worthy and unworthy. To practice metta meditation is to give the rarest gift: a gift that does not demand a return.

The Buddha describes how to practice this love in an early discourse called the “Karaniya Metta Sutta.”

A group of monks approach the Buddha complaining about the spirits living in the forest causing nearby villagers to suffer. The Buddha advises against fighting or driving them away. Instead, he encourages practicing boundless love toward them, wishing them happiness, peace and ease.

The monks do as recommended, practicing loving-kindness meditation for several weeks. Over time, noticing how happy the monks became, the spirits began to practice loving-kindness, too, because they also wanted to be happy. The practice changed the spirits’ behavior, and they stopped harassing the villagers.

How to practice loving-kindness

In the fifth century, a Sri Lankan monk named Buddhaghosa composed an important meditation text called the Visuddhimagga, or “The Path of Purification.” This text is sacred to Theravada Buddhists.

Buddhaghosa provides instructions for how to practice loving-kindness meditation. Contemporary teachers adapt and modify these instructions. However, the general format of this meditation tends to be consistent.

Loving-kindness meditation begins with a practice of mindfulness in order to calm the mind and body and to remember to come back to the now.

Next, this meditation involves softly reciting several traditional phrases and visualizing an audience who will receive loving-kindness as these words are spoken. The phrases are:

• May I/you/they/we be filled by loving-kindness

• May I/you/they/we be safe from inner and outer dangers.

• May I/you/they/we be well in body and mind.

• May I/you/they/we be at ease and happy.

Traditionally, the meditation starts with yourself – the pronoun will be “I.” Then, the meditation involves picturing a beloved person – and it does not even have to be a person; it can be a pet or an animal – and directing loving-kindness to them. The pronoun in the meditation will change to “you.”

After this, the meditation involves directing loving-kindness to a wider circle of friends and loved ones – the pronoun will change to “they.” Finally, the meditation involves gradually including more and more people in your well wishes: the folks in your community and town, people everywhere, animals and all living beings, and the whole Earth, and the pronoun will change to “we.”

Many versions of this meditation invite practitioners to express metta for people who have caused them difficulty, including to someone seen to be an “opponent.”

However, teachers including the Zen master, poet and peace activist Thich Nhat Hanh recommend practicing this type of metta meditation only once you are well established in directing loving-kindness at yourself and those you are close to.

Why practice loving-kindness meditation?

Clinical research shows that loving-kindness meditation has a positive effect on mental health. It could help lessen anxiety and depression, increase life satisfaction and improve self-acceptance; it could also reduce self-criticism.

There is also evidence that loving-kindness meditation increases a sense of connection. Practicing loving-kindness could increase happiness while strengthening feelings of kinship with all living beings, a few of the benefits of metta meditation described by the Buddha in the Karaniya Metta Sutta.

So if you’re feeling disconnected from others, ill at ease or just disenchanted with a holiday that has become overrun by capitalism on this Valentine’s Day, you might consider trying loving-kindness meditation.

Jeremy David Engels, Liberal Arts Endowed Professor of Communication, Penn State

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Lauren Blum, University of Colorado Boulder

There are trillions of charged particles – protons and electrons, the basic building blocks of matter – whizzing around above your head at any given time. These high-energy particles, which can travel at close to the speed of light, typically remain thousands of kilometers away from Earth, trapped there by the shape of Earth’s magnetic field.

Occasionally, though, an event happens that can jostle them out of place, sending electrons raining down into Earth’s atmosphere. These high-energy particles in space make up what are known as the Van Allen radiation belts, and their discovery was one of the first of the space age. A new study from my research team has found that electromagnetic waves generated by lightning can trigger these electron showers.

A brief history lesson

At the start of the space race in the 1950s, professor James Van Allen and his research team at the University of Iowa were tasked with building an experiment to fly on the United States’ very first satellite, Explorer 1. They designed sensors to study cosmic radiation, which is caused by high-energy particles originating from the Sun, the Milky Way galaxy, or beyond.

After Explorer 1 launched, though, they noticed that their instrument was detecting significantly higher levels of radiation than expected. Rather than measuring a distant source of radiation beyond our solar system, they appeared to be measuring a local and extremely intense source.

This measurement led to the discovery of the Van Allen radiation belts, two doughnut-shaped regions of high-energy electrons and ions encircling the planet.

Scientists believe that the inner radiation belt, peaking about 621 miles (1000 kilometers) from Earth, is composed of electrons and high-energy protons and is relatively stable over time.

The outer radiation belt, about three times farther away, is made up of high-energy electrons. This belt can be highly dynamic. Its location, density and energy content may vary significantly by the hour in response to solar activity.

The discovery of these high-radiation regions is not only an interesting story about the early days of the space race; it also serves as a reminder that many scientific discoveries have come about by happy accident.

It is a lesson for experimental scientists, myself included, to keep an open mind when analyzing and evaluating data. If the data doesn’t match our theories or expectations, those theories may need to be revisited.

Our curious observations

While I teach the history of the space race in a space policy course at the University of Colorado, Boulder, I rarely connect it to my own experience as a scientist researching Earth’s radiation belts. Or, at least, I didn’t until recently.

In a study led by Max Feinland, an undergraduate student in my research group, we stumbled upon some of our own unexpected observations of Earth’s radiation belts. Our findings have made us rethink our understanding of Earth’s inner radiation belt and the processes affecting it.

Originally, we set out to look for very rapid – sub-second – bursts of high-energy electrons entering the atmosphere from the outer radiation belt, where they are typically observed.

Many scientists believe that a type of electromagnetic wave known as “chorus” can knock these electrons out of position and send them toward the atmosphere. They’re called chorus waves due to their distinct chirping sound when listened to on a radio receiver.

Feinland developed an algorithm to search for these events in decades of measurements from the SAMPEX satellite. When he showed me a plot with the location of all the events he’d detected, we noticed a number of them were not where we expected. Some events mapped to the inner radiation belt rather than the outer belt.

This finding was curious for two reasons. For one, chorus waves aren’t prevalent in this region, so something else had to be shaking these electrons loose.

The other surprise was finding electrons this energetic in the inner radiation belt at all. Measurements from NASA’s Van Allen Probes mission prompted renewed interest in the inner radiation belt. Observations from the Van Allen Probes suggested that high-energy electrons are often not present in this inner radiation belt, at least not during the first few years of that mission, from 2012 to 2014.

Our observations now showed that, in fact, there are times that the inner belt contains high-energy electrons. How often this is true and under what conditions remain open questions to explore. These high-energy particles can damage spacecraft and harm humans in space, so researchers need to know when and where in space they are present to better design spacecraft.

Determining the culprit

One of the ways to disturb electrons in the inner radiation belt and kick them into Earth’s atmosphere actually begins in the atmosphere itself.

Lightning, the large electromagnetic discharges that light up the sky during thunderstorms, can actually generate electromagnetic waves known as lightning-generated whistlers.

These waves can then travel through the atmosphere out into space, where they interact with electrons in the inner radiation belt – much as chorus waves interact with electrons in the outer radiation belt.

To test whether lightning was behind our inner radiation belt detections, we looked back at the electron bursts and compared them with thunderstorm data. Some lightning activity seemed correlated with our electron events, but much of it was not.

Specifically, only lightning that occurred right after so-called geomagnetic storms resulted in the bursts of electrons we detected.

Geomagnetic storms are disturbances in the near-Earth space environment often caused by large eruptions on the Sun’s surface. This solar activity, if directed toward Earth, can produce what researchers term space weather. Space weather can result in stunning auroras, but it can also disrupt satellite and power grid operations.

We discovered that a combination of weather on Earth and weather in space produces the unique electron signatures we observed in our study. The solar activity disturbs Earth’s radiation belts and populates the inner belt with very high-energy electrons, then the lightning interacts with these electrons and creates the rapid bursts that we observed.

These results provide a nice reminder of the interconnected nature of Earth and space. They were also a welcome reminder to me of the often nonlinear process of scientific discovery.

Lauren Blum, Assistant Professor of Atmospheric and Space Physics, University of Colorado Boulder

This article is republished from The Conversation under a Creative Commons license. Read the original article.