Can Generative AI improve social science?

Generative AI that can produce realistic text, images, and other human-like outputs is currently transforming many different industries. Yet it is not yet known how such tools might influence social science research. I argue Generative AI has the potential to improve survey research, online experiments, automated content analyses, agent-based models, and other techniques commonly used to study human behavior. In the second section of this article, I discuss the many limitations of Generative. I examine how bias in the data used to train these tools can negatively impact social science research-as well as a range of other challenges related to ethics, replication, environmental impact, and the proliferation of low-quality research. I conclude by arguing that social scientists can address many of these limitations by creating open-source infrastructure for research on human behavior. Such infrastructure is not only necessary to ensure broad access to high-quality research tools, I argue, but also because the progress of AI will require deeper understanding of the social forces that guide human behavior.

What is bad news in fertility care? A qualitative analysis of staff and patients' accounts of bad and challenging news in fertility care.

STUDY QUESTION: What do fertility staff and patients think is bad news in fertility care? SUMMARY ANSWER: Staff and patients agree bad news is any news that makes patients less likely to achieve parenthood spontaneously or access and do successful treatment, but their appraisals of how bad the news is are differently influenced by specific news features and the context of its delivery. WHAT IS KNOWN ALREADY: Bad news is common in fertility care, but staff feel unprepared to share it and four in 10 patients react to it with unanticipated emotional or physical reactions. Research has paid much attention to how bad news should be shared, but considerably less to what news is perceived as bad, despite the fact this may dictate elements of its delivery. STUDY DESIGN, SIZE, DURATION: Two cross-sectional, online, mixed-method surveys (active 7 January-16 July 2022) were distributed to fertility staff and patients across the UK and Europe. PARTICIPANTS/MATERIALS, SETTING, METHODS: Staff inclusion criteria were being a healthcare professional working in fertility care and having experience of sharing bad news at least once a month. Patients' inclusion criteria were being adults and having had a conversation in which staff shared or explained bad news concerning their fertility care within the last 2 months. Surveys were created in English using Qualtrics, reviewed by patients and healthcare professionals, and distributed via social media, Prolific, fertility organizations, and scientific societies. Patients were asked, regarding the last time bad news were shared with them, 'What was the bad news?' and 'What other news would you consider bad news in fertility care?'. Staff were asked to 'List the three most challenging topics of bad news you share with your patients'. Staff and patient data were separately thematically analysed to produce basic codes, organized into sub-themes and themes. Themes emerging from patients' and staff data were compared and synthesized into meta themes. MAIN RESULTS AND THE ROLE OF CHANCE: Three hundred thirty-four staff accessed the survey, 286 consented, and 217 completed (65% completion rate). Three hundred forty-four patients accessed the survey, 304 consented, and 222 completed (64% completion rate). Eighty-five percent of participants were women, 62% resided in Europe, and 59% were in private care. Average staff age was 45.2 (SD = 12.0), 44% were embryologists or lab technicians, 40% were clinicians (doctors, consultants, or physicians), and 8% nurses or midwifes. Average patient age was 32.2 (SD = 6.4) and 54% had children. Staff answers originated 100 codes, 19 sub-themes and six themes. Patients' answers produced 196 codes, 34 sub-themes, and 7 themes. Staff and patient themes were integrated into three meta-themes reflecting main topics of bad news. These were Diagnosis and negative treatment events and outcomes, Inability to do (more) treatment, and Care and patient factors disrupting communication. Staff and patients agreed that some news features (uncertain, disruptive, definitive) made news more challenging but disagreed in relation to other features (e.g. unexpected/expected). Patient factors made bad news more challenging to staff (e.g. difficult emotions) and care factors made bad news more challenging to patients (e.g. disorganized care). LIMITATIONS, REASONS FOR CAUTION: Participants were self-selected, and most were women from private European clinics. Questions differed for staff and patients, focused on subjective perceptions of news, and did not measure news impact. WIDER IMPLICATIONS OF THE FINDINGS: The badness of fertility news is not only a product of the extent to which the news compromises parenthood goals but also of its features (timing, nature, number) and the context in which the news is delivered. Guidance on sharing bad news in fertility care needs to go beyond easing the process for patients to also consider staff experiences. Guidance may need to be tailored to news features and context. STUDY FUNDING/COMPETING INTEREST(S): Cardiff University funded the research. S.G., J.B., O'.H., and A.D. report funding from the Higher Education Funding Council for Wales and the European Society for Human Reproduction and Embryology (ESHRE) to develop fertiShare: a sharing bad news eLearning course for fertility care. fertiShare will be distributed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International Licence (CC BY-NC-SA 4.0). No other conflicts are reported in relation to this work. TRIAL REGISTRATION NUMBER: N/A.

Measuring non-events: infertility estimation using cross-sectional, population-based data from four countries in sub-Saharan Africa.

STUDY QUESTION: Does the prevalence of 12-month infertility in Burkina Faso, Côte d'Ivoire, Kenya, and Uganda differ between women trying to conceive and the broader population of women exposed to unprotected sex, and how are prevalence estimates affected by model assumptions and adjustments? SUMMARY ANSWER: Estimates of 12-month infertility among tryers ranged from 8% in Burkina Faso to 30% in Côte d'Ivoire, increasing substantially among a larger population of women exposed to unprotected intercourse. WHAT IS KNOWN ALREADY: While having a child is a fundamental human experience, the extent to which women and couples experience infertility is a neglected area of research, particularly in sub-Saharan Africa. Existing estimates of infertility in this region vary widely from 2% to 32%, however, potential impacts of variability in study populations and model assumptions have not been well-examined. STUDY DESIGN, SIZE, DURATION: We used cross-sectional nationally representative survey data from Burkina Faso, Côte d'Ivoire, Kenya, and Uganda. We employed a multi-stage cluster random sampling design with probability proportional to the size selection of clusters within each country to produce representative samples of women aged 15-49. Samples ranged from 3864 in Côte d'Ivoire to 9489 in Kenya. PARTICIPANTS/MATERIALS, SETTING, METHODS: We created two analytic samples in each country-tryers and a broader sample of women exposed to unprotected sex-exploring differences in population characteristics and estimating the period prevalence of 12-month infertility using the current duration (CD) approach. We also examined the impact of several model assumptions within each of the two analytic samples, including adjustments for recent injectable contraceptive use, unrecognized pregnancy, infertility treatment, underreported contraceptive use, and sexual activity. MAIN RESULTS AND THE ROLE OF CHANCE: Employing the CD approach among tryers produced an overall 12-month infertility prevalence of 7.9% (95% CI 6.6-12.7) in Burkina Faso, 29.6% (95% CI 15.3-100.0) in Côte d'Ivoire, 24.5% (95% CI 16.5-34.6) in Kenya, and 14.7% (95% CI 8.1-22.4) in Uganda. Results among women exposed to unprotected intercourse indicated much higher levels of infertility, ranging from 22.4% (95% CI 18.6-30.8) in Uganda to 63.7% (95% CI 48.8-87.9) in Côte d'Ivoire. Sensitivity analyses suggest infertility estimates are particularly sensitive to adjustments around pregnancy recognition timing and sexual activity, with little impact of adjustments for recent injectable contraceptive use, infertility treatment, and underreporting of traditional and coital dependent contraceptive use. LIMITATIONS, REASONS FOR CAUTION: There was substantial digit preference in responses at 12 months, particularly among the tryers, which could introduce bias. Data quality concerns in the reproductive calendar may impact the accuracy of the CD approach among the broader sample of women exposed to unprotected sex, particularly with regard to underreported contraceptive use, induced and spontaneous abortions, and unrecognized pregnancies. Lastly, we lacked information on postpartum amenorrhea or abstinence. WIDER IMPLICATIONS OF THE FINDINGS: Understanding the inconsistencies in definition and analytic approach and their implications for infertility estimation is important for reliably monitoring population-level infertility trends, identifying factors influencing infertility, improving prevention programs, and ensuring access to quality treatment and services. STUDY FUNDING/COMPETING INTEREST(S): This study was supported by grants from the Bill & Melinda Gates Foundation (INV009639) and the National Institute of Child Health and Human Development (K01HD107172). The funders were not involved in the study design, analyses, manuscript writing, or the decision to publish. The authors have no conflicts of interest to declare. TRIAL REGISTRATION NUMBER: N/A.

Recruiting people with selected citizenships for the health interview survey GEDA Fokus throughout Germany: evaluation of recruitment efforts and recommendations for future research.

BACKGROUND: Germany is the second most common country of immigration after the US. However, people with own or familial history of migration are not represented proportionately to the population within public health monitoring and reporting. To bridge this data gap and enable differentiated analyses on migration and health, we conducted the health interview survey GEDA Fokus among adults with Croatian, Italian, Polish, Syrian, or Turkish citizenship living throughout Germany. The aim of this paper is to evaluate the effects of recruitment efforts regarding participation and sample composition. METHODS: Data collection for this cross-sectional and multilingual survey took place between 11/2021 and 5/2022 utilizing a sequential mixed-mode design, including self-administered web- and paper-based questionnaires as well as face-to-face and telephone interviews. The gross sample (n = 33436; age range 18-79 years) was randomly drawn from the residents' registers in 120 primary sampling units based on citizenship. Outcome rates according to the American Association for Public Opinion Research, the sample composition throughout the multistage recruitment process, utilization of survey modes, and questionnaire languages are presented. RESULTS: Overall, 6038 persons participated, which corresponded to a response rate of 18.4% (range: 13.8% for Turkish citizenship to 23.9% for Syrian citizenship). Home visits accounted for the largest single increase in response. During recruitment, more female, older, as well as participants with lower levels of education and income took part in the survey. People with physical health problems and less favourable health behaviour more often took part in the survey at a later stage, while participants with symptoms of depression or anxiety more often participated early. Utilization of survey modes and questionnaire languages differed by sociodemographic and migration-related characteristics, e.g. participants aged 50 years and above more often used paper- than web-based questionnaires and those with a shorter duration of residence more often used a translated questionnaire. CONCLUSION: Multiple contact attempts, including home visits and different survey languages, as well as offering different modes of survey administration, increased response rates and most likely reduced non-response bias. In order to adequately represent and include the diversifying population in public health monitoring, national public health institutes should tailor survey designs to meet the needs of different population groups considered hard to survey to enable their survey participation.

Building physician wellness into the culture: evaluating a family physician well-being programme using the physician wellness inventory.

PURPOSE: Family physicians have a higher incidence of burnout, dissatisfaction, and disengagement compared to other medical specialties. Addressing burnout on the individual and systemic level is important to promoting wellness and preventing deleterious effects on physicians and patients. We used the Physician Wellness Inventory (PWI) to assess the effects of a wellness programme designed to equip family physicians with skills to address burnout. METHODS: The PWI is a fourteen-item 5-point Likert scale broken down into 3 scores; (i) career purpose, (ii) cognitive flexibility, and (iii) distress. The PWI was distributed to a cohort of n = 111 family physician scholars at 3 time points: January 2021, May-June 2021, and October 2021. The response rate was 96.4% at baseline, and 72.1% overall. Demographic information was collected to assess differences. The survey was distributed online through Qualtrics (Provo, UT). RESULTS: Cognitive Flexibility scores at the endpoint were higher for POC scholars than white scholars (P = 0.024). Distress scores for all groups decreased over time. Female scholars were more nervous, and anxious at the start than male scholars (P = 0.012), which decreased over time (P = 0.022). New career scholars were more likely than later career scholars to be distressed (P = 0.007), but both groups' distress decreased over time (P = 0.003). Later career scholars' feelings of being bothered by little interest or pleasure in doing things decreased more than new career scholars (endpoint: P = 0.022; overall: P = 0.023). CONCLUSIONS: The wellness programme shows improvement in PWI scores, indicating the programme content should be evaluated further for system level improvements.

Reproductive autonomy and the experience of later-than-desired pregnancy: results from a cross-sectional survey of reproductive-aged women in Uganda.

BACKGROUND: The focus of reproductive autonomy research has historically been on the experience of unintended pregnancy and use of contraceptive methods. However, this has led to the neglect of a different group of women who suffer from constraints on their reproductive autonomy-women who experience pregnancies later than they desire or who are unable to become pregnant. This study examines the extent of later-than-desired pregnancy among women and evaluates the sociodemographic and reproductive factors associated with this experience in Uganda. METHODS: We use data from the Performance Monitoring for Action Uganda 2022 female survey. We restricted the nationally representative sample of reproductive-aged women to those who were currently pregnant or who had ever given birth (n = 3311). We compared the characteristics of women across fertility intention categories (wanted pregnancy earlier, then, later, or not at all) of their current or most recent birth and used multivariable logistic regression to examine factors independently associated with having a pregnancy later than desired compared to at a desired time. RESULTS: Overall, 28.3% of women had a later-than-desired pregnancy. Nearly all sociodemographic and reproductive characteristics were associated with the desired pregnancy timing of women's current or most recent pregnancy. Having higher education [adjusted odds ratio (aOR) 2.41, 95% confidence interval (CI) 1.13-5.13], having sought care for difficulties getting pregnant (aOR 2.12, 95% CI 1.30-3.46), and having less than very good self-rated health (good health aOR 1.74, 95% CI 1.12-2.71; moderate health aOR 1.77, 95% CI 1.09-2.86; very bad health aOR 4.32, 95% CI 1.15-16.26) were all independently significantly associated with increased odds of having a later-than-desired pregnancy. Being nulliparous (aOR 1.98, 95% CI 0.99-3.95) was also borderline significantly associated with having a later-than-desired pregnancy. CONCLUSIONS: Identifying those who have later-than-desired pregnancies is essential if we seek to make progress towards supporting women and couples in achieving their reproductive goals, not just preventing pregnancies. Research on desired pregnancy timing in sub-Saharan Africa should be expanded to capture later-than-desired pregnancies, a population which is invisible in existing data. This work has public health implications due to commonalities in the factors associated with mistimed and unintended pregnancies and their link to poorer health and potentially poorer pregnancy outcomes.

Association between physical activity and allostatic load among pregnant women.

Allostatic load (AL) in pregnant women is associated with maternal and infant health outcomes. Whether physical activity (PA) is a modifiable factor associated with AL during pregnancy is unknown. In this cross-sectional study, including 725 pregnant women in 3 different trimesters, 8 biomarkers were included, and the high-risk quartile approach based on sample distribution was used to construct AL index (ALI). ALI <2 was defined as a low level and ≥2 as a high level. Student's t-test or Mann-Whitney U test and chi-squared test or Fisher exact test were used to compare differences in AL with different demographic characteristics among pregnant women. The relationship between PA and AL in pregnant women was analyzed using a binary logistic regression model. The results show that the detection rate of high-risk AL during pregnancy was 47.3%. In the adjusted model, sufficient PA was related to a lower AL than insufficient PA (OR = .693, 95%CI:.494,.971; p = .033). Compared with low- and high-intensity PAs, moderate-intensity PA was associated with lower AL (OR = .645, 95%CI:.447,.930; p = .019). The results suggest that PA is a modifiable factor related to AL, and intervention is recommended to be carried out in the first trimester to prevent the increased likelihood of high AL as pregnancy progresses. In addition, health care personnel should encourage pregnant women to participate in PA, especially moderate-intensity PA, in order to obtain lower AL and promote maternal and child health.

Perspectives on remote learning of orthotic fabrication by certified hand therapists.

BACKGROUND: The COVID-19 pandemic caused disruption to continuing educational opportunities for hand therapists. In response, some courses were offered via online platforms, including virtual orthotic fabrication courses. It is important to determine the effectiveness and benefits of these courses for educating certified hand therapists and examine if remote learning of orthotic fabrication skills has continued merit and relevance. PURPOSE: To investigate the value and effectiveness of orthotic fabrication courses taught in a virtual format. STUDY DESIGN: Cross-sectional, mixed methods survey study. METHODS: A 31-item survey consisting of Likert-type, direct response, and open-ended questions about experiences and opinions of virtual orthotics courses was electronically delivered to certified hand therapists. Data analysis included descriptive and correlational statistics to highlight frequencies, ranges, and relationships between the participant demographics and opinions/experiences. Thematic analysis guided the coding of the qualitative data. RESULTS: A total of 459 responded, with a response rate of 9.7%. Most respondents had not participated in online courses on orthotic fabrication. Those that did reported high satisfaction but noted that clinical experience and knowledge from previous courses influenced this experience. Most participants felt that novice clinicians and students would not gain enough skills and confidence from online courses. However, participants with all levels of experience found the courses to be of value. CONCLUSIONS: Results suggest that while online learning of this skill set is valuable and effective, it is most beneficial for experienced clinicians. Disadvantages included the lack of instructor feedback necessary for hands-on skill development and the lack of peer interaction. Advantages included convenience of time, cost, accessibility, and the ability to revisit the topic as needed. Online learning of orthotic fabrication skills is a sustainable option for clinicians seeking to advance their skills. Nevertheless, it is not a substitute for initial training for novice hand therapists due to the lack of feedback and skill development.

A sequential Moken scaling approach to evaluate response quality in survey research.

Careless responding, where participants do not fully engage with item content, is pervasive in survey research. Left undetected, carelessness can compromise the interpretation and use of survey results, including information about participant locations on the construct, item difficulty, and the psychometric quality of the instrument. We present and illustrate a sequential procedure for evaluating response quality in survey research using indicators from Mokken scale analysis (MSA). We use a real data illustration and a simulation study to compare a sequential procedure to a standalone procedure. We also consider how identifying and removing responses with evidence of poor measurement properties affects item quality indicators. Results suggest that the sequential procedure was effective in identifying potentially problematic response patterns that may not always be captured by traditional methods for identifying careless responders but was not always sensitive to specific carelessness patterns. We discuss implications for research and practice.

Modeling insufficient effort responses in mixed-worded scales.

Surveys often add reverse-coded questions to monitor respondents with insufficient effort responses (IERs) but often wrongly assume that all respondents consistently answer all questions with full effort. By contrast, this study expanded the mixture model for IERs and ran a simulation via LatentGOLD to show the harmful consequences of ignoring IERs to positively and negatively worded questions: less test reliability, bias and less accuracy in slope and intercept parameters. We showed its practical application to two public data sets: Machiavellianism (five-point scale) and self-reported depression (four-point scale).