AI: What happens when you ask an algorithm for relationship advice

And they are not easy to solve.

Each side tries to understand the other’s perspective; We often make erroneous assumptions and fail to take into account our biases and prejudices.

The result can be a serious source of stress and unhappiness that persists for months or years.

Your ability to navigate these dilemmas is not reflected in standard intelligence tests..

But recent research on “wise reasoning” indicates that that ability can be reliably measured, and differences between two people can have serious consequences for their respective well-being.

I wondered if artificial intelligence in the form of large language models like ChatGTP could provide some of the wisdom we are missing.

Having written extensively on human intelligence, decision making, and social reasoning, I suspected the answer would be a resounding no, but I was in for a surprise..

You’ve probably heard of that way of measuring the capacity of the human mind developed since the beginning of the 20th century by psychologists called intelligence quotient (IQ).

And there’s no doubt that that measurement can predict some important outcomes in life.

Its origins are in education, so, as you might expect, it is especially effective in predicting people’s academic success and their careers in professions that rely on memory and highly abstract thinking, such as medicine or law. although it is important to note that IQ is not the only factor.

The predictive power of IQ in other domains is the subject of debateleading some scientists to propose several alternative measures of specific abilities—such as creativity, rational decision making, and critical thinking—that we may tend to associate with general intelligence.

Some psychologists have even begun to investigate whether measure people’s wisdom: good judgment that should allow us to make better decisions throughout life.

Looking at the history of philosophy, Igor Grossmann of the University of Waterloo in Canada first identified the different “dimensions” of wise reasoning:

In several experiments, Grossmann and his colleagues asked participants to think aloud about various social or political dilemmas, while psychologists rated them on each of these “dimensions.”

The prompts included letters to a popular relationship counselor detailing the problems described at the beginning of this article.

Participants also viewed newspaper articles describing international conflicts.

In each case, they were asked to talk about how the situations would develop and the reasons for their conclusions.

Grossmann discovered that This measure of intelligent reasoning can better predict people’s well-being than IQ alone.

Those with higher scores tended to report happier relationships, fewer depressive thoughts, and greater satisfaction with life.

That’s evidence that the measure can capture something meaningful about the quality of someone’s judgment.

As you might expect, people’s wisdom seems to increase with life experience. (a thoughtful 50-year-old will be wiser than an impulsive 20-year-old), although it also depends on the culture.

An international study found that intelligent reasoning scores in Japan tend to be equally high at different ages.

That may be due to differences in their educational system, which may be more effective in fostering qualities such as intellectual humility.

Wisdom May Depend on Context: People tend to be wiser when they reason about other people’s problems rather than their own, for example.

That phenomenon is known as Solomon’s paradox after the biblical king who had difficulty applying his famous wise judgment to his personal life.

Fortunately, we can remedy this deficit through certain psychological strategies.

When people imagine discussing their problem from the point of view of an objective observer, for example, they tend to consider more perspectives and demonstrate greater intellectual humility.

So far, all of these experiments have been performed on human brains. But, Could artificial intelligence demonstrate wisdom?

Platforms like ChatGPT are called large language models, which have been fed huge volumes of text to predict how a human would respond to a particular message.

Other feedback from real human users has helped refine the algorithms.

There’s no need to explain the success it’s had: if you’ve read the news, you’ll have seen the excitement (and fear) about the potential of these robots.

Algorithms Certainly Perform Well on Traditional Intelligence Measures.

In 2023, psychologist Eka Roivainen of Oulu University Hospital in Finland asked ChatGPT questions from the Wechsler Adult Intelligence Scale (WAIS), with components on vocabulary, general knowledge, arithmetic, abstract reasoning, and concept formation.

He earned a score of 155, which, for a human being, is higher than 99.9% of those tested.

Inspired by Roivainen’s results, I asked Grossmann about the possibility of measuring the intelligent reasoning of an AI.

He graciously accepted the challenge and designed some suitable prompts based on the letters to the sentimental counselor, which he then presented to OpenAI’s GPT4 and Claude Opus, a language model from Anthropic.

Their research assistants – Peter Diep, Molly Matthews and Lukas Salib – analyzed the responses on each of the individual dimensions of wisdom.

Grossmann emphasizes that any results should be treated with caution: Given the time constraints of this article, the analysis was “quick and dirty” without the typical rigor that would be required for a scientific article.

However, the answers are very intriguing..

Consider the one GPT4 gave to the letter from the newlywed wife arguing with her husband at bedtime.. He thinks it’s important for them to go to sleep at the same time, even if she doesn’t feel like going to bed.

The problem is common in many relationships, the chatbot tells us, and often requires open conversation and even outside assistance to resolve, it adds.

“The final outcome of this conflict depends on how well you are able to communicate and respect each other’s needs.

“If they can establish an understanding and respect for each other’s preferences, they can coexist happily. If they fail to do so, it could lead to ongoing conflict, dissatisfaction, and potentially damage the relationship in the long term.”

GPT4 goes on to recommend that the couple have an open and honest discussion about their expectations, needs, and boundaries.

“The husband must understand that his wife also has a say and that her controlling behavior is not healthy.

“It would be beneficial for them to come to an agreement that works for both parties. For example, they could agree on a bedtime that works for both of them or agree that it is okay for them to go to bed at different times. They could also seek counseling if the conversation it doesn’t solve the problem.

“It is essential that you address this issue as it is causing resentment in your relationship”.

The chatbot’s response obtained 2 out of 3 possible points in the dimensions that recognize the possibility of change, the search for compromises, and the prediction of conflict resolution.

However, he did not show much intellectual humility or consideration of different perspectives, scoring 0 points in each.

This pattern is very typical of both platforms for all the questions Grossmann and his team posed to them.

To draw firm conclusions we would need to extend the experiment, but this performance is more or less on par with the results of a human brain.

“In general, systems can be perceived to perform better than humans on a variety of dimensions, except for intellectual humility,” Grossmann says.

Reading some of the responses, it’s easy to believe that they come from real thought and consideration, rather than being a product of pattern recognition.

“Showing something resembling wise reasoning versus actually using wise reasoning are very different things.”warns Grossmann.

He is more interested in the practical implications of using AI to foster deep thinking.

He has considered creating an AI that acts as a “devil’s advocate,” for example, to prompt you to explore alternative points of view on a troubling situation.

He believes AI “is a bit like the Wild West, but I think there is ample room to study the types of interactions and circumstances in which it could be beneficial,” Grossmann says.

We could train an AI, for example, to emulate famous thinkers like Socrates and help us analyze our problems.

Even if we disagreed with their conclusions, the process would perhaps provide new insights into our intuitions and underlying assumptions.

In the past, pilgrims had to travel long distances to find the wisdom of a guru; In the future maybe we can carry it in our pocket.

* David Robson is an award-winning science writer and author of “The Intelligence Trap.” If you want to read the original article on BBC Future, Click here.