The quest for happiness has puzzled humans for centuries. Researchers have long debated whether this elusive state comes from within, is shaped by external influences, or a combination of both. A recent study published in Nature Human Behaviour sheds new light on this complex issue, revealing that happiness can arise from either individual factors, societal circumstances, or both – and which path is true varies greatly across people.

“We need to understand the sources of happiness to develop effective interventions,” said Emorie Beck, assistant professor of psychology at the University of California, Davis, and lead author of the study. The researchers investigated two primary models of happiness: the “bottom-up” perspective, where overall well-being arises from satisfaction with life domains such as wealth, enjoyable work, and fulfilling relationships; and the “top-down” approach, where happiness is driven by personal attitudes and qualities, like mindfulness or therapy.

However, surveys have shown that only a portion of the happiness gap between groups can be attributed to factors like income and life expectancy. This suggests that individual differences in mental states play a significant role in determining overall well-being. A third model proposes that bottom-up and top-down influences interact with each other to generate overall happiness.

Beck and her coauthors analyzed data from over 40,000 individuals across five countries who participated in repeated surveys of life satisfaction. The findings revealed roughly equal groups demonstrating each pattern: some were influenced by individual factors (top-down), while others were shaped by external circumstances (bottom-up). In addition, a few individuals exhibited bidirectional influences, and some showed no clear connection between their personal and societal well-being.

The study’s results imply that measuring subjective wellbeing at the population level may not accurately reflect individual experiences. If policymakers aim to improve happiness across society, they must address both external factors like health, income, housing, and jobs, as well as individual qualities such as personal resilience and purpose in life.

Importantly, the most effective policies will be tailored to the individual themselves. Targeting external factors for individuals whose happiness is not determined by them would likely be ineffective.

“These things are treated separately, but they aren’t really,” Beck said. “They feed into each other at a personal level.” The work was supported in part by grants from the National Institute on Aging.

The University of East Anglia has conducted a study that reveals a crucial difference between essays written by students and those generated by AI tools like ChatGPT. While AI-generated essays are impressive in their coherence and grammatical soundness, they fall short when it comes to injecting a personal touch into their content. The researchers analyzed 145 essays written by real university students and another 145 generated by ChatGPT, comparing the two in terms of engagement markers – techniques that enhance clarity, connection, and persuade readers.

The study found that student-written essays consistently featured a rich array of engagement strategies, making them more interactive and persuasive. These included rhetorical questions, personal asides, and direct appeals to the reader. In contrast, ChatGPT-generated essays tended to be impersonal, mimicking academic writing conventions but lacking the conversational nuance and personal touch that human writers bring to their work.

The researchers attribute this difference to the nature of AI training data and statistical learning methods, which prioritize coherence over conversational nuance. This reflects a broader concern that relying too heavily on AI tools could lead to a decline in critical literacy and thinking skills among students.

Despite these findings, the study does not dismiss the potential role of AI in education. Instead, it suggests that tools like ChatGPT should be used as teaching aids rather than shortcuts. By harnessing the power of AI while preserving human engagement and creativity, educators can create a more balanced learning environment that benefits both students and teachers.

This research has significant implications for educators worldwide, particularly in spotting cheating and promoting critical literacy and ethical awareness in the digital age. The study’s findings highlight the importance of fostering personal touch and critical thinking skills in students, rather than relying solely on AI-generated content.

The faster planning system developed by MIT researchers uses machine learning to reduce solve time by up to 50 percent and produce a solution that better meets a user’s objective, such as on-time train departures. The new method could also be applied to other complex logistical problems like scheduling hospital staff or assigning airline crews.

Engineers often break down these kinds of problems into a sequence of overlapping subproblems that can each be solved in a feasible amount of time. However, the overlaps cause many decisions to be needlessly recomputed, making it take much longer to reach an optimal solution.

The researchers’ new approach learns which parts of each subproblem should remain unchanged and freeze those variables to avoid redundant computations. A traditional algorithmic solver then tackles the remaining variables.

“This is a very complex combinatorial scheduling problem,” says Cathy Wu, a member of the Laboratory for Information and Decision Systems at MIT. “Our approach can be applied without modification to all these different variants.”

The researchers’ technique, which they call learning-guided rolling horizon optimization (L-RHO), teaches a machine-learning model to predict which operations should be recomputed when the planning horizon rolls forward.

To test their approach, the researchers compared L-RHO to several base algorithmic solvers and specialized solvers. It outperformed them all, reducing solve time by 54 percent and improving solution quality by up to 21 percent.

Their method continued to outperform all baselines even when tested on more complex variants of the problem, such as factory machines breaking down or extra train congestion.

“Our approach can be applied without modification to all these different variants,” says Wu. “It even outperformed additional baselines we created to challenge our solver.”

L-RHO can also adapt if the objectives change, automatically generating a new algorithm to solve the problem – all it needs is a new training dataset.

In the future, the researchers want to better understand the logic behind their model’s decision to freeze some variables, but not others. They also want to integrate their approach into other types of complex optimization problems like inventory management or vehicle routing.

This work was supported by the National Science Foundation and MIT’s Research Support Committee, among others.