The 2024 strike at Samsung Electronics in South Korea was the first in the company’s fifty-five-year history. It sought to disrupt production first on less-automated foundry lines producing legacy semiconductors and then on highly automated lines producing high-bandwidth memory, stacked memory for artificial intelligence (AI) accelerators. Production effects remained contested, and stoppages yielded no clear bargaining gain. This article argues that capital investment and automation distribute labor power unevenly within a firm: differences in production systems’ age and degree of automation can separate sites of greater potential labor leverage from those of greater strategic value. I call this relationship among fixed-capital age, automation, and labor dependence “capital vintage.” At Samsung, newer capacity producing its most valuable chips was more automated, while older lines remained more labor-dependent but less strategically valuable. Capital vintage anchors the analysis of leverage, legibility, and legitimacy. The strike showed that potential production leverage did not automatically make wage demands publicly legible, especially when strikers were portrayed as among the country’s top earners. The union therefore gave greater weight to an existing occupational-health strategy, drawing on decades of civil-legal conflicts that made injured workers’ bodily harm a recognizable basis for claims against Samsung. Yet its 2025 internal crisis showed that public legibility could not substitute for representative legitimacy, which depended on union members’ support. The article brings workers who manufacture AI’s material foundations into scholarship on worker power and automation. It draws on Korean- and English-language daily, union and company documents, statutory instruments, and civil-society materials.

The growing use of advanced artificial intelligence (AI) in public decision-making raises a host of concerns. In addition to impacting the rights and interests of individuals, the delegation of decision-making by government actors to increasingly autonomous AI systems could undermine the public character of the resulting decisions and erode vital democratic processes. To unpack this concern, we liken the use of AI technology in public decision-making to the outsourcing of government functions to private actors. Using this analytic lens, we make three contributions. First, by delineating the nature and scope of genuinely public decisions, we establish conditions for the legitimate use of autonomous AI systems in public institutions. Second, drawing on a fine-grained technical analysis of contemporary AI technology, we illustrate that autonomous systems rarely satisfy the conditions of legitimacy in practice. Third, we explore avenues for better satisfying these conditions of public decision-making and propose uses of AI that protect and even enhance democratic institutions.

Systematic reviews are the gold standard for evidence synthesis, but the manual title and abstract screening of thousands of references creates a severe bottleneck. Existing automated tools have historically struggled to achieve the near-perfect recall (sensitivity) required for reliable reviews. We developed MechaScreener as a ‘zero-shot’ automated screening tool that utilises a large language model to rank article relevance. The tool requires no initial training data or manual pre-screening, as MechaScreener directly applies user-provided question elements (PICO) or inclusion/exclusion criteria to assign an ordinal relevance score (1–5) to each reference. We evaluated the tool in two phases: a development phase using five reference libraries to optimise prompts, and an independent evaluation phase using 10 diverse Cochrane review libraries (comprising both randomised controlled trials [RCTs] and non-RCTs) containing over 58,000 references. In the evaluation dataset, MechaScreener achieved a perfect recall of 1.00 (100%) in all 10 reviews, ensuring that no relevant articles were missed (pooled record-level 95% CI: 0.98–1.00). Concurrently, it achieved an overall macro-average specificity of 0.61 (61%, cluster bootstrap 95% CI: 0.46–0.75; pooled record-level specificity 0.59, 95% CI: 0.59–0.60). Specificity varied from 0.21 in broad public health topics to 0.91 in precise pharmacological interventions, reflecting the tool’s built-in conservatism when evaluating ambiguous abstracts. By safely eliminating over 60% of irrelevant literature during the initial screening phase without compromising recall, MechaScreener functions as a highly reliable but low-effort ‘first-pass’ filter, allowing researchers to substantially reduce manual workloads and reallocate resources towards full-text review and data extraction.

Evidence synthesis (ES) is costly and time-intensive. Artificial intelligence (AI) tools are increasingly used to automate or semi-automate ES tasks. We conducted a living rapid review of AI tools that automate or semi-automate ES tasks. We searched for published articles and preprints through MEDLINE, Embase, and the Association for Computing Machinery Digital Library from January 1, 2021 through October 15, 2025, supplemented by hand-searching. We included existing reviews to capture evaluations before 2021. Screening was conducted in duplicate at the title/abstract level, with single full-text screening and data extraction verified by a second reviewer. We included 134 studies. Performance varied by task and evaluation approach. Search automation tools showed poor recall (median 17%), although human-in-the-loop approaches improved performance. Tools for identifying randomized controlled trials demonstrated high median recall (98%) and precision (92%). Semi-automated abstract screening tools using active learning reduced screening burden by approximately 50% while maintaining 95% recall. Large language models demonstrated variable accuracy for abstract screening (median 88% recall), full-text screening (median 96% recall), and data extraction (median 63% accuracy). Risk of bias (RoB) assessment tools showed moderate agreement with human reviewers for the Cochrane RoB tool version 1 (median 71%) but performed less well with other RoB tools. AI tools can meaningfully reduce burden in study design identification and abstract screening but perform inconsistently across other evidence synthesis tasks. Tools for search automation, data extraction, and RoB assessment require further development. No existing tool can currently replace human expertise.

Automation-assisted title–abstract screening is routinely evaluated using metrics such as recall, precision, and workload reduction, and results are increasingly summarized across studies. Synthesis presupposes that evaluation results refer to comparable target quantities. This article examines whether screening evaluation results produced under diverse automation-assisted workflows can meaningfully support cumulative inference, and under what conditions cross-study comparison is warranted. This conceptual article treats screening evaluation targets as workflow-induced rather than task-inherent. Evaluation design is decomposed into four workflow dimensions (representation, inference, governance, and evaluation) plus the reference-decision set produced or specified. Recurring choice patterns across these elements define evaluation configurations. Two study-level configurations and one synthesis-level pattern are illustrated: endogenous adjudication, inherited adjudication, and aggregation of heterogeneous targets. Across configurations, evaluation targets are induced by workflow-specific reference-decision sets and, in some designs, model-based target domains rather than by a common reference set. Nominally similar metrics can therefore estimate different quantities across studies, and apparent cumulativeness can arise from pooling estimates of different quantities. Standard heterogeneity statistics cannot distinguish variation in method effectiveness from variation in what is being estimated. Comparison is warranted only when targets align before pooling. Differences across screening evaluations may reflect differences in what is being estimated rather than statistical heterogeneity around a common estimand. When evaluation targets are not aligned across the framework, aggregate summaries can describe reported results but cannot support cumulative inference. Improved aggregation requires both better study-level specification of evaluation targets and explicit attention to configuration variation at synthesis .

A key task in conducting systematic reviews is deduplicating the results from database searching. Deduplication using reference management software can be time-consuming and prone to error, while automated tools can be expensive and lack transparency. To support review teams, we evaluated eight deduplication tools: (1) The Automated Systematic Search Deduplicator (ASySD); (2) Covidence; (3) Deduklick; (4) EPPI-Reviewer; (5) PICO Portal; (6) Rayyan; (7) The Systematic Review Accelerator (SRA) Deduplicator: Focused; (8) The SRA Deduplicator: Relaxed. Five randomly selected Cochrane reviews had their searches rerun to create five gold standard sets. We compared the gold standard sets to the outputs of the eight deduplication tools and evaluated the results for: (1) unique records removed; (2) duplicate records retained; (3) time taken to deduplicate. Summed across all five reviews, the unique records removed in error ranged from 2 to 22. The three best tools were: (1) Rayyan; (2) Covidence; (3) SRA Deduplicator: Focused. The duplicate records retained in error ranged from 34 to 280, the three best tools were: (1) ASySD; (2) Rayyan; (3) EPPI-Reviewer. The time taken to deduplicate ranged from one minute to 20 hours and 34 minutes, the three fastest tools were: (1) SRA Deduplicator: Relaxed; (2) Deduklick; (3) Covidence. No tool performed so poorly that we don’t recommend using it. But, as all the tools had strengths and weaknesses, some are expensive while others require large amounts of manual checking time, we recommend review teams compare the tools across all three outcomes and choose the tool that best suits their needs.

The development of instrumentation, control and automation (ICA) in water operations during half a century is reviewed, and new challenges are described. The ideal ICA system contains a quality team of people who feel a deep sense of ownership of the system and who are committed to the continuous improvement ethics; an instrumentation system that gathers adequate process variable information; a monitoring system to gather data, process and display the data, detect and isolate measurement faults or process abnormal situations, assist in diagnosis and advice; a control system to meet the goals of the operation.

• • ICA is not one scientific discipline; it combines a multitude of scientific and engineering disciplines, here called a “decathlon” combination.

• • ICA is a hidden technology. It is ubiquitous in most industrial processes, including urban water systems, and reveals how processes are connected. When everything works as intended, it is not noted, but if things go wrong, it will be observed.

• • ICA in the water industry has about 50 years of history and is now well recognized.

• • Computers had become more affordable in the late 1960s. It was recognized that wastewater treatment systems are truly dynamic. All the 14 ICA conferences, from 1973 to 2025, have addressed all aspects of ICA methodology and implementations. The author has had the privilege to participate in all the 14 conferences.

• • Technology push and demand pull not only has led to more advanced operations. The rapid development of process knowledge, machine learning, AI, computing power and communication can realize operation also in a system-wide perspective.

• • There is an increasing demand for water reuse and circular management of water, and ICA has the potential to play an important role. Systems thinking, involving the complete urban water system cycle, is a great issue today. To succeed here, it is necessary to expand cooperation between problem owners, the water industry and methodology researchers in academia.

How does technological change affect social policy preferences? We advance this lively debate by focusing on the role of dual vocational education and training (VET). Existing literature would lead us to expect that dual VET increases demand for compensatory social policy and magnifies the effect of automation risk on such demands. In contrast, we contend that dual VET weakens demand for compensatory social policy through three non-mutually exclusive mechanisms that we refer to as (i) material self-interest; (ii) workplace socialization; and (iii) skill certification. We further hypothesize that dual VET mitigates the association between automation risk and social policy preferences. Analyzing cross-national individual data from the European Social Survey and national-level data on education systems, we find strong evidence for our argument. The paper advances the debate on social policy preferences in the age of automation and sheds new light on the relationship between skill formation and social policy preferences.

Administrative burden describes the learning costs, psychological costs, and compliance costs people face when attempting to interface with the government, particularly in seeking a benefit. Algorithmic and automated processes offer the potential of reducing administrative burdens, but scant empirical research has determined to what, if any effect. This study uses the case of criminal record expungement in two policy contexts: traditional, court petition-based systems and newly enacted automated systems, to understand if and how administrative burden persists, and whether and how these burdens operate differently in the context of the criminal legal system. Drawing on interviews with 105 expungement-eligible people, we find that while automated expungement schemes shift the burden from petitioner to state to initiate the process, automation inadvertently creates new administrative burdens via failure to notify, partial clearances, and opaque data processes. Furthermore, respondents described how automation failed to provide a sense of confirmation from the state that their sentence was truly completed, rehabilitation had been acknowledged, or that collateral consequences should no longer wield the same power. Overall, we argue that leveraging automation to reduce burdens must include information availability by design; otherwise policy reforms may fail to fully achieve their goals.

The exponential growth of scientific literature poses increasing challenges for evidence synthesis. Systematic reviews (SRs) usually rely on keyword-based database searches, which are limited by inconsistent terminology and indexing delays. Citation searching—identifying studies that cite or are cited by known relevant articles—offers a complementary route to uncover additional evidence but remains poorly automated and integrated into screening workflows. We developed BibliZap, an open-source, fully automated citation-searching tool built on Lens.org data, performing multi-level forward and backward citation searches with relevance-based ranking. Its performance was evaluated across 66 published SRs, comparing five approaches: (1) PubMed-only searches; (2) PubMed followed by BibliZap restricted to the top 500 ranked results; (3) PubMed followed by full BibliZap screening; and (4–5) two exploratory early-stop strategies where BibliZap was initiated after identifying the first or the first three PubMed relevant records. The primary outcome was sensitivity, with secondary assessments of screening workload and precision. When used after PubMed screening, BibliZap increased mean sensitivity from 75% to 97%, achieving complete recall in over half of the reviews. Screening only the top 500 outputs still allowed over 90% of reviews to reach or exceed 80% recall. BibliZap recovered a median of three additional included articles per review, not retrieved by PubMed, while adding a median of 6,450 additional records. Citation searching via BibliZap enhances the completeness of evidence retrieval in SRs based on restricted database searches and supports transparent, scalable workflows adaptable to rapid and exploratory review contexts.

In recent years, utilizing technologies, such as virtual reality in mental healthcare and treatment, has developed significantly. This study aimed to investigate the effect of using virtual reality (VR) technology on controlling anxiety and reducing fear of heights (acrophobia). This study was a randomized controlled trial conducted in Birjand, Iran, in 2020. 120 participants were recruited and randomly allocated into two groups: intervention and control. The intervention group underwent a single simulated exposure to height using a virtual reality headset. The Beck Anxiety Inventory, alongside a researcher-developed questionnaire were administered as pre-tests to assess acrophobia. Data analysis was performed using SPSS version 23, with significance level at 0.05. The intervention group showed significantly reduced anxiety and acrophobia scores immediately and 1 month after exposure (P < 0.05). Post-exposure, both anxiety and acrophobia scores were significantly lower in the intervention group compared to the control group (P = 0.03 and P < 0.001, respectively), with no significant differences between groups before exposure or 1 month later (P > 0.05). The study concluded that VR technology is an effective tool for reducing anxiety and acrophobia. This approach appears to hold significant promise as a therapeutic modality for psychiatrists treating patients with acrophobia.

We examine long‐run effects of automation risk on turnout. We expect gendered negative effects because men's turnout is more sensitive to job loss and earnings, but negative effects might be offset by populist right‐wing mobilization on economic grievances. We rely on population‐wide administrative data to avoid well‐known biases in survey data. We find both men and women with high automation risk to suffer in the labour market, but automation risk is associated with lower turnout for men only. The negative association with turnout is weaker where the populist right is stronger, consistent with mobilization on economic grievances. Finally, we show experimentally that priming of automation risk produces null findings, suggesting that risks need to have material consequences to affect political behaviour. Our findings imply that technological change has contributed to the emergence of gender gaps in turnout and populist voting as well as the participation drop among the working class.