Boosting: The Science of Fostering Better Decisions

Rethinking how to Measure and Improve Food Decisions
Boosting More Meaningful Use of Digital Technologies With Self-Nudges
Policy Paper: Promoting Self-Regulatory Competences in Children and Adolescents
Boosting Human Competences with Interpretable and Explainable Artificial Intelligence
Boosting Risk Communication With Experiential Simulations

Everyone struggles to make good choices—what to eat, where to invest, whom to trust. The typical story is that people struggle with decisions like these because human beings are simply not rational enough to do what's in their own interest. Policymakers, buying into this view, have focused on nudging citizens in the "right" direction rather than empowering them to figure things out for themselves. 

ARC's work on boosting offers a fresh alternative. Instead of subtly steering people's behavior, boosts aim to enhance people's competences to help them make the right call in all areas of life, leaving them better equipped to reckon with risks, eat healthily, manage their money, protect the environment, and navigate the traps of the digital world.

Boosts matter because lifelong learning, self-efficacy, and agency matter. People need to keep adapting to meet new challenges, but often lack the time, energy, or opportunity to do so. Simple, practical tools that enable people to cope with rapidly changing environments are urgently needed—all the more so in the face of powerful industries like Big Tech and Big Food, which seek to exploit humans' weak spots for profit. Boosts empower people to unmask industry tricks, fight back against manipulation, and make choices that align with their own goals and values.

An upcoming book (Hertwig & Grüne-Yanoff, in press) and a recent review in Annual Review of Psychology (Herzog & Hertwig, 2025) argue that to meet the grand challenges of the time—including climate change, manipulation online, threats to democracy, international migration, aging populations, and child safety online—it is crucial to recognize and foster the human potential to learn, to comprehend, and to adapt. This does not mean that other changes, especially systematic changes such as smart regulations are not necessary. But regulations and laws need to be complemented by smart citizens.

Boosting is a call to rethink behavioral science: to shift away from steering people and toward empowering them to steer themselves (Hertwig et al., 2025). It invites policymakers, educators, and citizens to champion competence, agency, and autonomy as the foundation of a more resilient future.

Key References

Rethinking how to Measure and Improve Food Decisions

A widely cited claim holds that people make over 200 "mindless" food decisions a day. The number has permeated public discourse, casting people as passive, mindless eaters who are unaware of their own choices. Claassen et al. (2025) found that this figure is essentially an illusion of precision. When people estimated their daily food decisions in one shot, they reported about 14. When the same question was unpacked into subcategories—what, when, where, how much, with whom—and the answers summed, the estimate ballooned to 227. This is a textbook subadditivity effect: Breaking a category into components increases frequency judgments. Add more subcategories and the number climbs further. The 200-decisions figure says more about the psychology of estimation than about actual eating behavior.

A randomized trial by Dallacker et al. (2023) suggests that people are not mindless eaters. In the study, 50 parent–child dyads ate two evening meals: one that lasted as long as their usual duration and one about 10 minutes longer. During longer meals, children ate roughly seven more bite-sized pieces of fruits and vegetables (equivalent to one extra 100 g serving, e.g., one medium apple). Crucially, children did not eat more of everything. They ate more of the bite-sized, ready-to-grab fruits and vegetables—but not more bread and cold cuts, which required assembly. The physical structure of the food shaped which items benefited from extra time. This points to a design insight that connects both papers: What counts as a "food decision" is not just a measurement question but a lever for intervention. Precisely operationalizing the decision unit—individual bites, coded from video—is what revealed this selectivity. A cruder measure would have missed it entirely.

Together, these findings reframe the challenge: The problem is not mindless eaters who need correcting. Rather, the issue is rushed meals and a lack of fruits and vegetables—more broadly, impoverished decision environments that need redesigning. Families with more time and accessible healthy foods shifted toward better choices on their own.

These insights into the design of food environments also informed a broader policy effort: a publication by the Berlin-Brandenburg Academy of Sciences and Humanities (Grune et al., 2025), in which an interdisciplinary group of experts—including ARC Director Ralph Hertwig—proposed policy recommendations for an urgently needed nutrition transition in Germany, emphasizing changes to food environments alongside individual behavior.

Key References

Boosting More Meaningful Use of Digital Technologies With Self-Nudges

Over the last decade, concerns have grown about the impact of social media use on people's well-being, attention, and self-control, in particular for heavy users and adolescents. Many existing interventions merely aim to reduce screen time, stopping shy of helping people use social media in an intentional, healthy manner. These interventions also rely heavily on individual willpower and paternalistic "one size fits all" restrictions that people often disregard. Could lightweight, user-chosen interventions that interrupt use at the moment it occurs encourage more intentional social media behavior over time and at scale?

Grüning et al. (2023) evaluated One Sec, a self-nudging app that introduces a brief interruption when users attempt to open selected social media apps (Figure 1). One Sec combines short friction with the explicit option to dismiss the app. In a 6-week field experiment (N = 280), it reduced target app openings by 57%—a result driven by both frequent dismissals and fewer opening attempts over time. A preregistered follow-up experiment (N = 500) showed that the ability to dismiss the app was the most effective intervention mechanism; messages encouraging deliberation alone had no effect.

A second large-scale policy experiment was conducted by the Danish Competition and Consumer Authority (2025) in collaboration with ARC's David Grüning, encompassing 269 adolescents aged 13 to 17 years. Small in-app interventions reduced daily social media use by 31–36%, halved social media sessions during school hours, and increased sleep duration by half an hour without increasing the fear of missing out. The effects were strongest among heavy users and individuals with low self-control, demonstrating the scalability and policy relevance of minimal, autonomy-preserving interventions. Taken together, these studies highlight the promise of digitally scaffolded self-nudging to boost more meaningful use of digital technologies.

Key References

Danish Competition and Consumer Authority. (2025). Disrupting social media habits: A field experiment with young Danish consumers. Link

Policy Paper: Promoting Self-Regulatory Competences in Children and Adolescents

Self-regulation is a crucial competence for today's children and teenagers, who are relentlessly confronted with highly engineered consumer environments that are designed to seize their attention and their money. A German National Academy of Sciences Leopoldina policy paper, written by an interdisciplinary group of experts including ARC Director Ralph Hertwig, argues that self-regulation should be systematically taught in schools.

Children's and adolescents' self-regulation competences—cognitive, emotional, motivational, and social abilities such as organizing tasks, directing attention, regulating emotions, resolving conflicts constructively, and participating in decisions—are foundational for their health, education, social lives, and general well-being. Current evidence paints a concerning picture of rising psychological burdens, physical risks from unhealthy diets and inactivity, and declining competences in mathematics, reading, and science, all compounded by risk factors including parental mental illness, low socioeconomic status, experiences of flight and migration, violence, and bullying. Digital and social media create opportunities such as building social networks and exploring and expressing one's political voice but they also present substantial risks.

State and society are obliged, under the United Nations Convention on the Rights of the Child, the Constitution of the World Health Organization, and the Basic Law of Germany, to secure young people's well-being and developmental opportunities. Failure carries high social and economic costs, including the burdens of mental illness and unrealized educational potential. Promoting self-regulation should be anchored as a guiding perspective across the German education system. Structural conditions must also improve, via measures such as providing stronger socioeconomic support for families and single parents, improving staffing and environments in early childhood centers and schools, and combatting industry practices that shape children's preferences and behavior (e.g., social media design, food marketing, tobacco and alcohol promotion).

Self-regulation develops through the interplay of genetics and environments. Early childhood is strongly shaped by families; eventually, early childhood centers and schools also play important roles. Research has identified effective approaches: supportive learning environments that combine cognitive activation with constructive support, mental health literacy that emphasizes self-regulation as learnable, and evidence-based methods (behavioral and cognitive–behavioral, mindfulness- and compassion-based, and body-oriented approaches). Digital technologies can reinforce these programs. Success requires qualified educational staff, adequate resources, and strong coordination between the education and healthcare sectors.

Recommended measures:
●      Make self-regulation a system-wide priority by embedding it in standards, curricula, teacher education and professional development, and institutional concepts; engage parents' councils, school conferences, unions, and associations.
●      Measure and monitor self-regulation with clear indicators integrated into national education monitoring and school improvement; strengthen data on mental health and well-being; use innovative, everyday assessments (e.g., smartphone-based real-time surveys); extend monitoring to preschoolers over time.
●      Implement evidence-based practice at scale: Integrate self-regulation into everyday pedagogy (classroom management, cognitive activation, constructive support) and establish shared, institution-wide concepts.
●      Provide tailored offerings for vulnerable groups.
●      Ensure continuous improvement and sustainability through regular evaluation and co-development with research institutions to secure scalability and long-term maintenance.

Key Reference

Deutsche Akademie der Naturforscher Leopoldina. (2025). Promoting the self-regulation skills of children and adolescents in day care centres and schools (Series on Science-Based Policy Advice: Position Paper). https://doi.org/10.26164/leopoldina_03_01273

Boosting Human Competences with Interpretable and Explainable Artificial Intelligence

Most artificial intelligence (AI) systems used in high-stakes settings, such as medical diagnostics and criminal justice, are opaque black boxes. The field of explainable AI (XAI) tries to address this by providing post hoc explanations of what these systems do. But there is a fundamental problem: Explanations cannot faithfully represent what the original model computes. If they could, the explanation would simply be the model, and there would be no need for the black box in the first place.

A promising alternative to black box AI is simple, inherently interpretable models, such as short decision trees or checklists. Particularly for the structured data common in medical, legal, and policy settings, these simple models can often match the accuracy of complex AI systems while being transparent by design.

In a conceptual analysis, Herzog and Franklin (2024) connect the distinction between interpretable and explainable AI to the boosting framework and ask: Can interpretable AI and XAI strengthen two kinds of AI-related human competences? The first competence is the ability to critically evaluate AI systems—for instance, detecting when a model relies on objectionable features (e.g., race) or has learned nonsensical relationships from biased training data. Interpretable models make such problems visible; with post hoc explanations, there is no guarantee that what users see reflects what the model actually does. The second competence concerns people's ability to make accurate judgments on their own, even after AI support is removed. Drawing on research on multiple-cue learning, the authors argue that interpretable models function much like "task information" interventions: By revealing how a few key features relate to outcomes, these models help people internalize useful decision strategies.

The paper proposes an experimental paradigm for testing these conjectures (Figure 2): Participants complete a judgment task without AI support, then work with different forms of AI assistance, and finally perform the task again unaided—revealing which forms of AI support produce competence gains that persist once the AI is taken away.

The upshot is that inherently interpretable models should be prioritized wherever feasible—not just because they sidestep the faithfulness problems of post hoc explanations, but because they hold the greatest promise for empowering human decision-makers. This opens a new chapter for the boosting approach: If the right AI systems can build competences that persist, AI need not only assist human decision-making—it can systematically strengthen it.

Key Reference

Boosting Risk Communication With Experiential Simulations

When communities face threats such as pandemics or extreme weather, public health officials typically respond with written warnings, statistical summaries, and data visualizations. These approaches assume that factual knowledge about risks is sufficient to motivate informed action. Yet consider what it might mean to truly understand the risks of smoking: not merely knowing that nicotine is addictive, but actually feeling the grip of craving and withdrawal. Traditional risk communication cannot convey the visceral reality that makes abstract dangers concrete.

Tiede et al. (2026) proposed a taxonomy that distinguishes the standard dimensions of risk information (i.e., possibility and likelihood of outcomes) from oft-neglected experiential dimensions. Simulated experiences—through digital interfaces, virtual reality environments, or embodied tools such as age-simulation suits—can communicate these overlooked aspects. Early evidence is promising: For instance, virtual reality simulations of consequences of climate change have increased people's increased risk perceptions and negative emotions. These responses, in turn, were associated with reduced intentions to consume dairy and meat.

The authors call for a research agenda that identifies which experiential dimensions matter most for different risks, clarifies the mechanisms through which simulations shape understanding, and develops ethical guidelines for balancing informative and potentially manipulate effects.

Key Reference