Scientific discovery in the age of artificial intelligence.

Artificial intelligence (AI) is being increasingly integrated into scientific discovery to augment and accelerate research, helping scientists to generate hypotheses, design experiments, collect and interpret large datasets, and gain insights that might not have been possible using traditional scientific methods alone. Here we examine breakthroughs over the past decade that include self-supervised learning, which allows models to be trained on vast amounts of unlabelled data, and geometric deep learning, which leverages knowledge about the structure of scientific data to enhance model accuracy and efficiency. Generative AI methods can create designs, such as small-molecule drugs and proteins, by analysing diverse data modalities, including images and sequences. We discuss how these methods can help scientists throughout the scientific process and the central issues that remain despite such advances. Both developers and users of AI toolsneed a better understanding of when such approaches need improvement, and challenges posed by poor data quality and stewardship remain. These issues cut across scientific disciplines and require developing foundational algorithmic approaches that can contribute to scientific understanding or acquire it autonomously, making them critical areas of focus for AI innovation.

Augmenting astrophysical scaling relations with machine learning: Application to reducing the Sunyaev-Zeldovich flux-mass scatter.

Complex astrophysical systems often exhibit low-scatter relations between observable properties (e.g., luminosity, velocity dispersion, oscillation period). These scaling relations illuminate the underlying physics, and can provide observational tools for estimating masses and distances. Machine learning can provide a fast and systematic way to search for new scaling relations (or for simple extensions to existing relations) in abstract high-dimensional parameter spaces. We use a machine learning tool called symbolic regression (SR), which models patterns in a dataset in the form of analytic equations. We focus on the Sunyaev-Zeldovich flux-cluster mass relation (YSZ - M), the scatter in which affects inference of cosmological parameters from cluster abundance data. Using SR on the data from the IllustrisTNG hydrodynamical simulation, we find a new proxy for cluster mass which combines YSZ and concentration of ionized gas (cgas): M ∝ Yconc3/5 ≡ YSZ3/5(1 - A cgas). Yconc reduces the scatter in the predicted M by ∼20 - 30% for large clusters (M ≳ 1014 h-1 M⊙), as compared to using just YSZ. We show that the dependence on cgas is linked to cores of clusters exhibiting larger scatter than their outskirts. Finally, we test Yconc on clusters from CAMELS simulations and show that Yconc is robust against variations in cosmology, subgrid physics, and cosmic variance. Our results and methodology can be useful for accurate multiwavelength cluster mass estimation from upcoming CMB and X-ray surveys like ACT, SO, eROSITA and CMB-S4.

A forward modeling approach to analyzing galaxy clustering with SimBIG.

We present cosmological constraints from a simulation-based inference (SBI) analysis of galaxy clustering from the SimBIG forward modeling framework. SimBIG leverages the predictive power of high-fidelity simulations and provides an inference framework that can extract cosmological information on small nonlinear scales. In this work, we apply SimBIG to the Baryon Oscillation Spectroscopic Survey (BOSS) CMASS galaxy sample and analyze the power spectrum, [Formula: see text], to [Formula: see text]. We construct 20,000 simulated galaxy samples using our forward model, which is based on 2,000 high-resolution Quijote[Formula: see text]-body simulations and includes detailed survey realism for a more complete treatment of observational systematics. We then conduct SBI by training normalizing flows using the simulated samples and infer the posterior distribution of [Formula: see text]CDM cosmological parameters: [Formula: see text]. We derive significant constraints on [Formula: see text] and [Formula: see text], which are consistent with previous works. Our constraint on [Formula: see text] is 27% more precise than standard [Formula: see text] analyses because we exploit additional cosmological information on nonlinear scales beyond the limit of current analytic models, [Formula: see text]. This improvement is equivalent to the statistical gain expected from a standard [Formula: see text] analysis of galaxy sample [Formula: see text]60% larger than CMASS. While we focus on [Formula: see text] in this work for validation and comparison to the literature, SimBIG provides a framework for analyzing galaxy clustering using any summary statistic. We expect further improvements on cosmological constraints from subsequent SimBIG analyses of summary statistics beyond [Formula: see text].

A Bayesian neural network predicts the dissolution of compact planetary systems.

We introduce a Bayesian neural network model that can accurately predict not only if, but also when a compact planetary system with three or more planets will go unstable. Our model, trained directly from short N-body time series of raw orbital elements, is more than two orders of magnitude more accurate at predicting instability times than analytical estimators, while also reducing the bias of existing machine learning algorithms by nearly a factor of three. Despite being trained on compact resonant and near-resonant three-planet configurations, the model demonstrates robust generalization to both nonresonant and higher multiplicity configurations, in the latter case outperforming models fit to that specific set of integrations. The model computes instability estimates up to [Formula: see text] times faster than a numerical integrator, and unlike previous efforts provides confidence intervals on its predictions. Our inference model is publicly available in the SPOCK (https://github.com/dtamayo/spock) package, with training code open sourced (https://github.com/MilesCranmer/bnn_chaos_model).

Predicting the long-term stability of compact multiplanet systems.

We combine analytical understanding of resonant dynamics in two-planet systems with machine-learning techniques to train a model capable of robustly classifying stability in compact multiplanet systems over long timescales of [Formula: see text] orbits. Our Stability of Planetary Orbital Configurations Klassifier (SPOCK) predicts stability using physically motivated summary statistics measured in integrations of the first [Formula: see text] orbits, thus achieving speed-ups of up to [Formula: see text] over full simulations. This computationally opens up the stability-constrained characterization of multiplanet systems. Our model, trained on ∼100,000 three-planet systems sampled at discrete resonances, generalizes both to a sample spanning a continuous period-ratio range, as well as to a large five-planet sample with qualitatively different configurations to our training dataset. Our approach significantly outperforms previous methods based on systems' angular momentum deficit, chaos indicators, and parametrized fits to numerical integrations. We use SPOCK to constrain the free eccentricities between the inner and outer pairs of planets in the Kepler-431 system of three approximately Earth-sized planets to both be below 0.05. Our stability analysis provides significantly stronger eccentricity constraints than currently achievable through either radial velocity or transit-duration measurements for small planets and within a factor of a few of systems that exhibit transit-timing variations (TTVs). Given that current exoplanet-detection strategies now rarely allow for strong TTV constraints [S. Hadden, T. Barclay, M. J. Payne, M. J. Holman, Astrophys. J. 158, 146 (2019)], SPOCK enables a powerful complementary method for precisely characterizing compact multiplanet systems. We publicly release SPOCK for community use.

Learning to predict the cosmological structure formation.

Matter evolved under the influence of gravity from minuscule density fluctuations. Nonperturbative structure formed hierarchically over all scales and developed non-Gaussian features in the Universe, known as the cosmic web. To fully understand the structure formation of the Universe is one of the holy grails of modern astrophysics. Astrophysicists survey large volumes of the Universe and use a large ensemble of computer simulations to compare with the observed data to extract the full information of our own Universe. However, to evolve billions of particles over billions of years, even with the simplest physics, is a daunting task. We build a deep neural network, the Deep Density Displacement Model ([Formula: see text]), which learns from a set of prerun numerical simulations, to predict the nonlinear large-scale structure of the Universe with the Zel'dovich Approximation (ZA), an analytical approximation based on perturbation theory, as the input. Our extensive analysis demonstrates that [Formula: see text] outperforms the second-order perturbation theory (2LPT), the commonly used fast-approximate simulation method, in predicting cosmic structure in the nonlinear regime. We also show that [Formula: see text] is able to accurately extrapolate far beyond its training data and predict structure formation for significantly different cosmological parameters. Our study proves that deep learning is a practical and accurate alternative to approximate 3D simulations of the gravitational structure formation of the Universe.

Fake news, real risks: How online discussion and sources of fact-check influence public risk perceptions toward nuclear energy.

This study seeks to understand how online discussion, fact-checking, and sources of fact-checks will influence individuals' risk perceptions toward nuclear energy when they are exposed to fake news. Using a 2 × 3 experimental design, 320 participants were randomly assigned to one of the six experimental conditions. Results showed an interaction effect between online discussion and exposure to fact-checking, in which online discussion lowered individuals' risk perception toward nuclear energy when a fact-check was unavailable. Of those who participated in the online discussion, those who viewed a fact-check posted by traditional media have higher risk perception as compared to those who viewed a fact-check posted by a fact-check organization. Our findings indicate that different fact-checking sources can have differential effects on public risk perceptions, depending on whether online discussion is involved. To curb the spread of fake news, different fact-checking strategies will need to be deployed depending on the situation.

Using the Marked Power Spectrum to Detect the Signature of Neutrinos in Large-Scale Structure.

Cosmological neutrinos have their greatest influence in voids: These are the regions with the highest neutrino to dark matter density ratios. The marked power spectrum can be used to emphasize low-density regions over high-density regions and, therefore, is potentially much more sensitive than the power spectrum to the effects of neutrino masses. Using 22 000 N-body simulations from the Quijote suite, we quantify the information content in the marked power spectrum of the matter field and show that it outperforms the standard power spectrum by setting constraints improved by a factor larger than 2 on all cosmological parameters. The combination of marked and standard power spectra allows us to place a 4.3σ constraint on the minimum sum of the neutrino masses with a volume equal to 1 (Gpc h^{-1})^{3} and without cosmic microwave background priors. Combinations of different marked power spectra yield a 6σ constraint within the same conditions.

Promoting Healthier Eating via Parental Communication: Development and Validation of the Active and Restrictive Parental Guidance Questionnaire (PARQ).

Parents are important sources of influence in the development of healthy eating among children and adolescents. Besides gatekeeping and modeling, parents serve as health educators and promoters, using intentional and persuasive communication to encourage healthier eating preferences and behaviors in children. Despite this, a lack of reliable and valid measures has limited the research on how parent-driven interpersonal communication about foods influence child food consumption outcomes. Building on the research in parental mediation of media consumption, and parenting practices in public health nutrition, this study details the development and validation of the active and restrictive parental guidance questionnaire with a sample of 246 children and adolescents at the scale development phase and another sample of 1,113 children and adolescents at the scale validation phase. Findings show that parents employ four communicative strategies to encourage a healthier diet: active guidance, general discussion, preventive restrictive guidance, and promotive restrictive guidance. The new measure was shown to have good validity and measurement model fit. Implications for future research are discussed.

Examining Organizational, Cultural, and Individual-Level Factors Related to Workplace Safety and Health: A Systematic Review and Metric Analysis.

One major gap in existing health communication research is that few studies have synthesized findings from the literature to map out what are the key factors related to workplace (a) safety awareness, (b) safety risks, (c) health awareness, and (d) health risks. This study bridges the gap by systematically reviewing what these organizational, cultural, and individual-level factors are, and examine the impact of workplace safety and health publications using traditional and alternative metrics in academic and non-academic settings. Through an iterative process of coding, the results revealed six categories of organizational (management commitment, management support, organizational safety communication, safety management systems, physical work environment, and organizational environment), two cultural (interpersonal support and organizational culture), and four individual-level (perception, motivation, attitude, and behavior) factors. In terms of impact, articles that were most impactful in academia (e.g., high citation count) may not necessarily receive the same amount of online attention from the public. Theoretical and practical implications for health communication were discussed.

Factors affecting engagement between academic faculty and decision-makers: learnings and priorities for a school of public health.

BACKGROUND: Schools of public health (SPHs) are increasingly being recognised as important contributors of human, social and intellectual capital relevant to health policy and decision-making. Few studies within the implementation science literature have systematically examined knowledge exchange experiences within this specific organisational context. The purpose of this study was therefore to elicit whether documented facilitators and barriers to engaging with government decision-makers resonates within an academic SPH context. We sought to understand the variations in such experiences at four different levels of government decision-making. Furthermore, we sought to elicit intervention priorities as identified by faculty. METHODS: Between May and December 2016, 211 (34%) of 627 eligible full-time faculty across one SPH in the United States of America participated in a survey on engagement with decision-makers at the city, state, federal and global government levels. Surveys were administered face-to-face or via Skype. Descriptive data as well as tests of association and logistic regression analyses were conducted using STATA. RESULTS: Over three-quarters of respondents identified colleagues with ties to decision-makers, institutional affiliation and conducting policy-relevant research as the highest facilitators. Several identified time constraints, academic incentives and financial support as important contributors to engagement. Faculty characteristics, such as research areas of expertise, career track and faculty rank, were found to be statistically significantly associated with facilitators. The top three intervention priorities that emerged were (1) creating incentives for engagement, (2) providing funding for engagement and (3) inculcating an institutional culture around engagement. CONCLUSIONS: The data suggest that five principal categories of factors - individual characteristics, institutional environment, relational dynamics, research focus and funder policies - affect the willingness and ability of academic faculty to engage with government decision-makers. This study suggests that SPHs could enhance the relevance of their role in health policy decision-making by (1) periodically measuring engagement with decision-makers; (2) enhancing individual capacity in knowledge translation and communication, taking faculty characteristics into account; (3) institutionalising a culture that supports policies and practices for engagement in decision-making processes; and (4) creating a strategy to expand and nurture trusted, relevant networks and relationships with decision-makers.

The influence of parental practices on child promotive and preventive food consumption behaviors: a systematic review and meta-analysis.

BACKGROUND: The family is an important social context where children learn and adopt eating behaviors. Specifically, parents play the role of health promoters, role models, and educators in the lives of children, influencing their food cognitions and choices. This study attempts to systematically review empirical studies examining the influence of parents on child food consumption behavior in two contexts: one promotive in nature (e.g., healthy food), and the other preventive in nature (e.g., unhealthy food). METHODS: From a total of 6,448 titles extracted from Web of Science, ERIC, PsycINFO and PubMED, seventy eight studies met the inclusion criteria for a systematic review, while thirty seven articles contained requisite statistical information for meta-analysis. The parental variables extracted include active guidance/education, restrictive guidance/rule-making, availability, accessibility, modeling, pressure to eat, rewarding food consumption, rewarding with verbal praise, and using food as reward. The food consumption behaviors examined include fruits and vegetables consumption, sugar-sweetened beverages, and snack consumption. RESULTS: Results indicate that availability (Healthy: r = .24, p < .001; Unhealthy: r = .34, p < .001) and parental modeling effects (Healthy: r = .32, p < .001; Unhealthy: r = .35, p < .001) show the strongest associations with both healthy and unhealthy food consumption. In addition, the efficacy of some parenting practices might be dependent on the food consumption context and the age of the child. For healthy foods, active guidance/education might be more effective (r = .15, p < .001). For unhealthy foods, restrictive guidance/rule-making might be more effective (r = -.11, p < .01). For children 7 and older, restrictive guidance/rule-making could be more effective in preventing unhealthy eating (r = - .20, p < .05). For children 6 and younger, rewarding with verbal praise can be more effective in promoting healthy eating (r = .26, p < .001) and in preventing unhealthy eating (r = - .08, p < .01). CONCLUSIONS: This study illustrates that a number of parental behaviors are strong correlates of child food consumption behavior. More importantly, this study highlights 3 main areas in parental influence of child food consumption that are understudied: (1) active guidance/education, (2) psychosocial mediators, and (3) moderating influence of general parenting styles.

The Augmented Cognitive Mediation Model: Examining Antecedents of Factual and Structural Breast Cancer Knowledge Among Singaporean Women.

As knowledge acquisition is an important component of health communication research, this study examines factors associated with Singaporean women's breast cancer knowledge using an augmented cognitive mediation model. We conducted a nationally representative study that surveyed 802 women between the ages of 30 and 70 using random-digit dialing. The results supported the augmented cognitive mediation model, which proposes the inclusion of risk perception as a motivator of health information seeking and structural knowledge as an additional knowledge dimension. There was adequate support for the hypothesized paths in the model. Risk perception was positively associated with attention to newspaper, television, Internet, and interpersonal communication. Attention to the three media channels was associated with interpersonal communication, but only newspaper and television attention were associated with elaboration. Interpersonal communication was positively associated with structural knowledge, whereas elaboration was associated with both factual and structural knowledge. Differential indirect effects between media attention and knowledge dimensions via interpersonal communication and elaboration were found. Theoretical and practical implications are discussed.

Effects of Exergaming and Message Framing in School Environments on Physical Activity Attitudes and Intentions of Children and Adolescents.

Although interventions targeting the health of students in schools are becoming common, few studies have examined how health messages operate at the group level in school environments. This study examines the effects of message-based health interventions (extrinsic vs. intrinsic goal framing) in group environments (exergame competitive vs. exergame noncompetitive) on eliciting attitudes and intentions toward physical activity among children and adolescents. We conducted a 7-week school-based intervention program involving 336 children and 259 adolescents in Singapore in which pre- and post-intervention responses were recorded. Our findings revealed the difference in responses between child and adolescent groups. Children who participated in noncompetitive exergames with extrinsically framed health messages and those who participated in competitive exergames with intrinsically framed health messages demonstrated more favorable attitudes toward physical activity. However, the same effects were absent in our adolescent group. These findings suggest that the integration of exergames into competitive and noncompetitive environments can serve as a gateway to traditional physical activity in schools when strategically combined with intrinsically and extrinsically framed messages. Practical and theoretical implications for schools and health educators are discussed.

For Fit's Sake: A Norms-Based Approach to Healthy Behaviors Through Influence of Presumed Media Influence.

Based on the influence of presumed media influence (IPMI) model as the theoretical framework, this study examines how injunctive norms and personal norms mediate the influence of healthy lifestyle media messages on public intentions to engage in two types of healthy lifestyle behaviors-physical activity and healthy diet. Nationally representative data collected from 1,055 adults in Singapore demonstrate partial support for the key hypotheses that make up the extended IPMI model, highlighting the importance of a norms-based approach in health communication. Our results indicate that perceived media influence on others indirectly shaped public intentions to engage in healthy lifestyle behaviors through personal norms and attitude, providing partial theoretical support for the extended IPMI model. Practical implications for health communicators in designing health campaigns media messages to motivate the public to engage in healthy lifestyle are discussed.

Value predispositions as perceptual filters: Comparing of public attitudes toward nanotechnology in the United States and Singapore.

This study compares public attitudes toward nanotechnology in the United States and Singapore, using large-scale survey data in both countries. Results indicate that Singaporeans tend to be more knowledgeable about and familiar with nanotechnology than the U.S. public. Singaporeans tend to perceive greater benefits and fewer potential risks of nanotechnology, and to indicate greater support for government funding for nanotechnology than the U.S. public. Between the two countries, perceived familiarity with nanotechnology and the benefits and risks of the emerging technology tend to be interpreted differently through the lens of value predispositions (religiosity and deference to scientific authority) and therefore they indirectly affect public support. Specifically, the U.S. public tends to use religiosity to interpret benefits and Singaporeans are inclined to use religiosity to think about risks. Deference to scientific authority also moderates the impact of perceived familiarity with nanotechnology on funding support for the technology among the U.S. public.

Examining how presumed media influence affects social norms and adolescents' attitudes and drinking behavior intentions in rural Thailand.

This study uses the influence of presumed media influence model as the theoretical framework to examine how perceived social norms (i.e., descriptive, subjective, and injunctive norms) will mediate the influence of pro- and antidrinking media messages on adolescents' intention to consume alcohol in rural Thailand. Data collected from 1,028 high school students indicate that different mechanisms underlie drinking intentions between nondrinkers and those who have consumed alcohol or currently drink. Among nondrinkers, perceived peer attention to prodrinking messages indirectly influenced adolescents' prodrinking attitudes and intentions to consume alcohol through all three types of perceived social norms. Among drinkers, perceived peer attention to pro- and antidrinking messages indirectly influenced adolescents' prodrinking attitudes and intentions to drink alcohol through perceived subjective norm. The findings provide support for the extended influence of presumed media influence model and have practical implications for how antidrinking campaigns targeted at teenagers in Thailand might be designed.

Factors related to employers' intent to hire, retain and accommodate cancer survivors: the Singapore perspective.

PURPOSE: Despite the growing importance of cancer and return-to-work issues in occupational rehabilitation literature in the last decade, academic discussion is largely limited to survivors' perspectives and some exploratory studies from the employer side. This paper applies two classic theoretical models-Theory of Planned Behavior and Social Cognitive Theory-and key measures from previous studies to identify explicit relationships that explain employer factors to hire and retain cancer survivors. METHODS: Data were collected from online surveys with senior management executives and senior human resource specialists from various organizations in Singapore, with a total of 145 responses. The 72-item survey instrument included a series of independent variables: (1) Attitudes toward cancer and cancer survivors; (2) Employers' efficacy; (3) Perceived moral obligation; (4) Employers' experience; (5) Outcome expectations; (6) Employment situation; (7) Social norms; and (8) Incentives, and dependent variables: (a) Employers' intention to hire cancer survivors; and (b) Employers' intention to retain cancer survivors. RESULTS: Regression analyses showed that the top three factors related to employers' intention to retain cancer survivors are perceived moral obligations (ß = .39, p < .001), followed by attitudes toward cancer (ß = .25, p < .01), and employment situation (ß = .17, p < .05). Employers' efficacy was associated with intention to hire (ß = .22, p < .05), coupled with attitude toward cancer survivors (ß = .22, p < .01). The findings also indicated the important role of existing relationship between an employer and an employee when it comes to retaining cancer survivors and government incentives for hiring cancer survivors in the workforce. CONCLUSIONS: The present study provided an avenue to implement the proposed model-a potential study framework for the management of cancer survivors at work. Findings revealed that different messages should be tailored to employers toward hiring and retention issues and provided useful guidelines for employer education materials.

Does knowledge make a difference? Understanding how the lay public and experts assess the credibility of information on novel foods.

Drawing on Metzger's dual-processing model of credibility assessment, this study examines how individuals with varying topical knowledge (laypersons vs experts) assess the credibility of information on novel foods. Online focus group discussions reveal that both groups share similar motivations for assessing the credibility of information on novel foods (e.g. personal relevance and concerns about the impact of unverified information on others). However, they differ in the barriers they encounter during the assessment of information credibility. Both groups employ analytical (e.g. evaluating content quality) and intuitive methods (e.g. looking at source credibility) to assess the credibility of novel food-related information. However, they differ in the cues used for credibility assessment. Laypersons tend to rely on superficial heuristics (e.g. social endorsement cues or surface features), whereas experts rely more on content features and scientific knowledge to evaluate information credibility. Theoretical and practical implications are discussed.