Catalysphere Articles https://catalysphere.co.jp/en/articles/ Research notes on AI, data, behavioral science, and organizations. en Stop counting AI agreement as review evidence https://catalysphere.co.jp/en/articles/llm-sycophancy-decision-review/ https://catalysphere.co.jp/en/articles/llm-sycophancy-decision-review/ Fri, 25 Sep 2026 15:00:00 GMT LLMs can shift answers toward a user's stance or pushback, and ratings miss it. Investment reviews should not count AI agreement as independent verification, should re-verify AI objections, and should probe how models update at rollout and on change. More audit logs do not make agent actions reconstructable https://catalysphere.co.jp/en/articles/ai-agent-audit-log-minimization/ https://catalysphere.co.jp/en/articles/ai-agent-audit-log-minimization/ Wed, 26 Aug 2026 15:00:00 GMT Design evidence for AI-agent actions around a chain of approval, execution, and state change—not full-content retention. Capture disputed content selectively under protection, then test reconstructability and minimization as separate release gates. AI transparency does not end with a single disclosure https://catalysphere.co.jp/en/articles/ai-transparency-disclosure-design/ https://catalysphere.co.jp/en/articles/ai-transparency-disclosure-design/ Wed, 26 Aug 2026 15:00:00 GMT A role-and-output reading of EU AI Act Article 50, translating human-facing disclosure, machine-readable marking, exceptions, and version evidence into one placement record. Reusing a Human Login Cannot Govern an Agent's Authority https://catalysphere.co.jp/en/articles/ai-agent-delegated-authorization/ https://catalysphere.co.jp/en/articles/ai-agent-delegated-authorization/ Wed, 26 Aug 2026 15:00:00 GMT Instead of reusing a human session, separate the principal from the agent and evaluate a delegation that binds resource, action, audience and lifetime at execution time. Interviews Suggested About 90% Less Time. More Capacity? https://catalysphere.co.jp/en/articles/ai-automation-capacity-conversion/ https://catalysphere.co.jp/en/articles/ai-automation-capacity-conversion/ Thu, 06 Aug 2026 15:00:00 GMT After AI-enabled RPA was introduced, interviews suggested monthly research time fell about 90%. Later, only 10–20% of estimated freed time was explained as new work or extra output. This note shows how to measure the time actually freed and where it went. Rethinking RPA: Execution Choices and Operational Handoffs in the AI Era https://catalysphere.co.jp/en/articles/rpa-redesign-operational-handoff/ https://catalysphere.co.jp/en/articles/rpa-redesign-operational-handoff/ Sun, 26 Jul 2026 15:00:00 GMT After work has been redesigned, decide how remaining steps combine core systems, APIs, RPA, AI, and people—and what proves the operation can run without its builder. How Far Should Workflows Be Redesigned After AI Adoption? https://catalysphere.co.jp/en/articles/ai-workflow-redesign/ https://catalysphere.co.jp/en/articles/ai-workflow-redesign/ Sat, 18 Jul 2026 15:00:00 GMT Treat faster local tasks as only one signal; redesign bottlenecks, exceptions, expert handoffs, and the allocation of released capacity across the full workflow. AI use-case selection depends on reallocating validation capacity, not a fixed ranking https://catalysphere.co.jp/en/articles/ai-use-case-portfolio-prioritization/ https://catalysphere.co.jp/en/articles/ai-use-case-portfolio-prioritization/ Sat, 18 Jul 2026 15:00:00 GMT Allocate exploratory and validation capacity using testability, harm, reversibility, and learning value—not expected impact alone. A Pilot Comparison Design for Moving Generative AI into Production https://catalysphere.co.jp/en/articles/ai-pilot-experiment-design/ https://catalysphere.co.jp/en/articles/ai-pilot-experiment-design/ Sat, 18 Jul 2026 15:00:00 GMT Compare the existing and AI-assisted workflows under aligned conditions, including human review and revision time, confirmed quality, and unfinished cases. Use a record to decide deployment, hold, retest, or stop. Is consent to AI as easy to refuse as it is to accept? https://catalysphere.co.jp/en/articles/dark-patterns-ai-consent/ https://catalysphere.co.jp/en/articles/dark-patterns-ai-consent/ Wed, 15 Jul 2026 15:00:00 GMT Based on evidence that button emphasis, obstruction, defaults, and repeated requests alter choices, audit the procedures and visibility of acceptance, refusal, and withdrawal symmetrically. Build a team where AI errors can be reported with confidence. https://catalysphere.co.jp/en/articles/psychological-safety-ai-teams/ https://catalysphere.co.jp/en/articles/psychological-safety-ai-teams/ Wed, 15 Jul 2026 15:00:00 GMT Do not mistake general research on psychological safety for AI-specific causal evidence; integrate dissent, error reporting, stop proposals, and retaliation checks into the execution procedures for AI implementation. Increasing the number of warnings does not necessarily increase safety. https://catalysphere.co.jp/en/articles/warning-fatigue-ai-controls/ https://catalysphere.co.jp/en/articles/warning-fatigue-ai-controls/ Wed, 15 Jul 2026 15:00:00 GMT Given habituation to repeated warnings, false alarms, and ignoring due to irrelevant notifications, design a warning budget by risk, display variations, and mandatory stops. Not whether nudges work, but how much they shrink https://catalysphere.co.jp/en/articles/nudge-scale-evidence/ https://catalysphere.co.jp/en/articles/nudge-scale-evidence/ Wed, 15 Jul 2026 15:00:00 GMT Based on meta-analysis corrections, re-analysis of publication bias, and large-scale administrative nudge trials, we shift the decision framework to prioritize baseline rates, absolute differences, heterogeneity, and costs in advance. Even if you entrust your thinking to AI, can you return to it on your own? https://catalysphere.co.jp/en/articles/cognitive-offloading-ai/ https://catalysphere.co.jp/en/articles/cognitive-offloading-ai/ Wed, 15 Jul 2026 15:00:00 GMT We distinguish between load reduction via external memory and the reduction of internal encoding and retrieval, measuring both the performance when AI is usable and the independent execution when it is not. Does the label "AI-generated work" evaluate the work itself? https://catalysphere.co.jp/en/articles/ai-label-creative-evaluation/ https://catalysphere.co.jp/en/articles/ai-label-creative-evaluation/ Wed, 15 Jul 2026 15:00:00 GMT From experiments showing that evaluations change when a work is labeled as AI-generated, we separate blind content evaluation from eligibility reviews of provenance, rights, and production process. Procedures other than outcomes are required for fair AI judgments. https://catalysphere.co.jp/en/articles/algorithmic-procedural-justice/ https://catalysphere.co.jp/en/articles/algorithmic-procedural-justice/ Wed, 15 Jul 2026 15:00:00 GMT We design procedures that do not equate fairness with merely displaying explanations, but instead enable individuals to understand, present their circumstances, correct data, and proceed to human review and relief. Do not limit recommended objectives solely to duration of stay. https://catalysphere.co.jp/en/articles/recommender-systems-wellbeing/ https://catalysphere.co.jp/en/articles/recommender-systems-wellbeing/ Wed, 15 Jul 2026 15:00:00 GMT Evidence that ranking methods yield inconsistent attitudinal effects despite altering exposure and usage necessitates the creation of a ranked evaluation separating information diversity, reliability, time utilization, and well-being. Do AI companions reduce loneliness or replace it? https://catalysphere.co.jp/en/articles/ai-companions-loneliness/ https://catalysphere.co.jp/en/articles/ai-companions-loneliness/ Wed, 15 Jul 2026 15:00:00 GMT Distinguish between connection immediately after conversation, loneliness after repeated use, comparison with humans, and relationship replacement; do not extrapolate short-term effects to long-term well-being. Is it rational to stop using AI after observing its errors? https://catalysphere.co.jp/en/articles/algorithm-aversion-control/ https://catalysphere.co.jp/en/articles/algorithm-aversion-control/ Wed, 15 Jul 2026 15:00:00 GMT Based on experiments showing that avoidance after observing algorithmic errors and acceptance with minor correction rights vary, we design operations that are neither mandatory use nor laissez-faire. Does making AI more human-like lead to appropriate trust? https://catalysphere.co.jp/en/articles/ai-anthropomorphism-trust/ https://catalysphere.co.jp/en/articles/ai-anthropomorphism-trust/ Wed, 15 Jul 2026 15:00:00 GMT Do not treat names, voices, appearances, and emotional expressions as uniform trust-enhancing measures; design them based on context, user emotions, understanding of capabilities, expectation violations, and handover to humans. The knowledge base of RAG becomes a searchable attack surface. https://catalysphere.co.jp/en/articles/rag-poisoning-defense/ https://catalysphere.co.jp/en/articles/rag-poisoning-defense/ Wed, 15 Jul 2026 15:00:00 GMT We mitigate the risk of malicious documents entering the top search results and guiding answers through write history, index monitoring, contradiction detection, post-generation verification, and isolation with re-indexing. Gathering any number of AIs does not result in independent wisdom. https://catalysphere.co.jp/en/articles/multi-agent-debate-reliability/ https://catalysphere.co.jp/en/articles/multi-agent-debate-reliability/ Wed, 15 Jul 2026 15:00:00 GMT Do not trust multi-LLM discussions based on vote counts; instead, determine the added value by assessing the diversity of initial answers, conformity, problem drift, retention of correct answers, and the difference from simple voting. Does correcting an LLM's knowledge break other capabilities? https://catalysphere.co.jp/en/articles/model-editing-side-effects/ https://catalysphere.co.jp/en/articles/model-editing-side-effects/ Wed, 15 Jul 2026 15:00:00 GMT We do not judge success solely by the target correct answer for knowledge editing; instead, we measure paraphrase, locality, general capabilities, and forgetting after continuous editing to establish a boundary for switching to RAG or retraining. How can one prove that AI has forgotten? https://catalysphere.co.jp/en/articles/machine-unlearning-verification/ https://catalysphere.co.jp/en/articles/machine-unlearning-verification/ Wed, 15 Jul 2026 15:00:00 GMT Do not treat the completion of a deletion API as proof of forgetting; instead, comprehensively verify target memory, retention utility, leakage, and continuous deletion by using a retrained model as the baseline. The average score of multilingual AI conceals inequities across languages. https://catalysphere.co.jp/en/articles/multilingual-llm-equity/ https://catalysphere.co.jp/en/articles/multilingual-llm-equity/ Wed, 15 Jul 2026 15:00:00 GMT Do not deploy based solely on the overall score of translation benchmarks; determine the scope of provision by analyzing culture-dependent items, native-speaker evaluations, worst-case performance per language, and relief pathways. Code written by AI must be reviewed for speed and safety separately. https://catalysphere.co.jp/en/articles/ai-code-security-review/ https://catalysphere.co.jp/en/articles/ai-code-security-review/ Wed, 15 Jul 2026 15:00:00 GMT Do not equate the productivity of generated code with safety; instead, treat vulnerabilities, user safety perception, static and dynamic testing, and independent review as gates for adopting high-impact code. The ability not to answer, included in LLM quality https://catalysphere.co.jp/en/articles/llm-abstention-uncertainty/ https://catalysphere.co.jp/en/articles/llm-abstention-uncertainty/ Wed, 15 Jul 2026 15:00:00 GMT Rather than prompting guesses solely based on accuracy, design the boundaries between answering, confirming, and stopping by accounting for semantic uncertainty, unanswerable questions, and the losses from incorrect answers versus abstention. Does that benchmark score measure unseen capabilities? https://catalysphere.co.jp/en/articles/benchmark-contamination-detection/ https://catalysphere.co.jp/en/articles/benchmark-contamination-detection/ Wed, 15 Jul 2026 15:00:00 GMT We examine the inclusion of the public test set in training data from multiple perspectives, including data duplication, completion behavior, and example order, to isolate suspicious scores from decision-making. When using an LLM as a judge, who audits the judge? https://catalysphere.co.jp/en/articles/llm-as-judge-bias/ https://catalysphere.co.jp/en/articles/llm-as-judge-bias/ Wed, 15 Jul 2026 15:00:00 GMT We establish boundaries to return to human evaluation by auditing not only the agreement rate with human evaluators but also through reversal of answer order, redundancy, self-preference, and difficult cases. Do not approve long-context AI by its nominal limit—set the effective range from position, evidence, and cost. https://catalysphere.co.jp/en/articles/long-context-reliability/ https://catalysphere.co.jp/en/articles/long-context-reliability/ Wed, 15 Jul 2026 15:00:00 GMT For owners approving long-document AI, this article compares full text, chunking, and retrieval under changing positions, task types, and noise, then sets an automation boundary from worst-case quality and total verification cost. Does algorithmic management deprive workers of discretion? https://catalysphere.co.jp/en/articles/algorithmic-management-worker-agency/ https://catalysphere.co.jp/en/articles/algorithmic-management-worker-agency/ Tue, 14 Jul 2026 15:00:00 GMT We organize work design by separating allocation, monitoring, evaluation, and reward by function, and setting minimum standards for autonomy, connectivity pressure, anxiety, and the ability to appeal, in addition to efficiency. Conversational-AI corrections must be measured through later memory and behavior https://catalysphere.co.jp/en/articles/ai-misinformation-correction-design/ https://catalysphere.co.jp/en/articles/ai-misinformation-correction-design/ Tue, 14 Jul 2026 15:00:00 GMT We present a correction protocol anchored to primary sources, integrating the potential of personalized correction dialogues, the LLM's own epistemic limits, the persistence of effects after warnings, and the decay of effectiveness. AI-written empathy does not replace human relational responsibility https://catalysphere.co.jp/en/articles/ai-mediated-empathy-authenticity/ https://catalysphere.co.jp/en/articles/ai-mediated-empathy-authenticity/ Tue, 14 Jul 2026 15:00:00 GMT We design an operation that distinguishes the quality of AI-generated empathetic text from the recipient's perceived understanding, connection, and source preference, while safeguarding human editorial responsibility and recipient choice. AI reasoning traces are investigative hypotheses, not audit trails https://catalysphere.co.jp/en/articles/chain-of-thought-auditability/ https://catalysphere.co.jp/en/articles/chain-of-thought-auditability/ Tue, 14 Jul 2026 15:00:00 GMT This article demonstrates an audit design that distinguishes between the fluency of Chain-of-Thought and internal judgment reproduction, focusing on answer accuracy, causal fidelity, hidden cues, and separated execution logs. AI-text detectors cannot stand alone as evidence for adverse decisions https://catalysphere.co.jp/en/articles/ai-generated-text-detection-evidence/ https://catalysphere.co.jp/en/articles/ai-generated-text-detection-evidence/ Tue, 14 Jul 2026 15:00:00 GMT We design an evidentiary procedure that distinguishes AI text detectors from generation-time watermarks and includes paraphrase resistance, unknown models, group-based false positives, the author's version history, and avenues for appeal. Recursive training on synthetic data can lose distribution tails before the average collapses https://catalysphere.co.jp/en/articles/synthetic-training-data-model-collapse/ https://catalysphere.co.jp/en/articles/synthetic-training-data-model-collapse/ Tue, 14 Jul 2026 15:00:00 GMT Instead of uniformly prohibiting the reuse of synthetic data, we design learning data operations that separate the replacement, accumulation, and selection of real data to preserve rare cases and diversity. How much should conversational AI be allowed to change human choices? https://catalysphere.co.jp/en/articles/ai-conversational-persuasion-autonomy/ https://catalysphere.co.jp/en/articles/ai-conversational-persuasion-autonomy/ Sun, 12 Jul 2026 15:00:00 GMT Personalized conversational AI can correct misconceptions or steer users toward commercial interests without their awareness. We define operational conditions that distinguish legitimate behavioral support from manipulation and show how to measure autonomy. More explanation does not automatically calibrate reliance on AI https://catalysphere.co.jp/en/articles/ai-reliance-calibration-behavioral-science/ https://catalysphere.co.jp/en/articles/ai-reliance-calibration-behavioral-science/ Sun, 12 Jul 2026 15:00:00 GMT Whether AI explanations, confidence displays, judgment order, and user control can truly reconcile the adoption of correct advice with the rejection of incorrect advice is examined by organizing conditions from behavioral experiments. AI coaching should be judged by behavior after support ends https://catalysphere.co.jp/en/articles/ai-behavior-change-coaching/ https://catalysphere.co.jp/en/articles/ai-behavior-change-coaching/ Sun, 12 Jul 2026 15:00:00 GMT Conversation continuity and lasting real-world behavior are distinct. Drawing on meta-analyses and field RCTs of conversational coaches, this article outlines evaluation designs that avoid mistaking short-term responses for behavior change. Data Minimization in the AI Era: Do Not Collect, Do Not Retain, Do Not Overlap https://catalysphere.co.jp/en/articles/ai-privacy-data-minimization/ https://catalysphere.co.jp/en/articles/ai-privacy-data-minimization/ Sun, 12 Jul 2026 15:00:00 GMT We do not leave AI privacy measures solely to post-collection anonymization; we reduce necessity, retention, duplication, learning, logging, and access across the entire lifecycle. Production models age even when the code does not change https://catalysphere.co.jp/en/articles/model-data-drift-monitoring/ https://catalysphere.co.jp/en/articles/model-data-drift-monitoring/ Sun, 12 Jul 2026 15:00:00 GMT Monitoring only changes in input distribution does not reveal actual harm; design multi-layered monitoring and response procedures that connect data, predictions, decisions, and delayed outcomes. Synthetic Data Is Not Anonymized: Designing Privacy and Utility https://catalysphere.co.jp/en/articles/synthetic-data-privacy-utility/ https://catalysphere.co.jp/en/articles/synthetic-data-privacy-utility/ Sun, 12 Jul 2026 15:00:00 GMT The mere format of artificial generation does not guarantee safety; we address attackers, memorization, differential privacy, and utility by use case in a single evaluation table. Why KPIs Break: Operationalizing Goodhart’s Law https://catalysphere.co.jp/en/articles/kpi-goodharts-law/ https://catalysphere.co.jp/en/articles/kpi-goodharts-law/ Sun, 12 Jul 2026 15:00:00 GMT We do not end the discussion with the notion that “metrics become targets and the frontline deteriorates”; instead, we distinguish bias toward easily measurable work, gaming, and actual improvement to design robust operations that resist breakdown. How much can human prediction improve through training, aggregation, and falsification? https://catalysphere.co.jp/en/articles/forecasting-human-judgment/ https://catalysphere.co.jp/en/articles/forecasting-human-judgment/ Sun, 12 Jul 2026 15:00:00 GMT Rather than framing predictive ability as individual talent, we consider practical conditions derived from probabilistic expression, brief training, aggregation of independent judgments, and continuous scoring and re-analysis. Design-first practice for causal thinking without RCTs https://catalysphere.co.jp/en/articles/causal-inference-without-rct/ https://catalysphere.co.jp/en/articles/causal-inference-without-rct/ Sun, 12 Jul 2026 15:00:00 GMT Begin causal inference from observational data not by choosing methods, but by designing comparisons; clarify the roles and limitations of target trials, propensity scores, and regression discontinuity. Where do data quality defects become decision-making errors? https://catalysphere.co.jp/en/articles/data-quality-decision-errors/ https://catalysphere.co.jp/en/articles/data-quality-decision-errors/ Sun, 12 Jul 2026 15:00:00 GMT By tracking data quality not only in terms of accuracy but also meaning, proxy indicators, and on-site correction pathways, we identify where numbers lead to judgment errors. Internal AI answers fail on knowledge lifecycle before search accuracy https://catalysphere.co.jp/en/articles/knowledge-base-content-operations/ https://catalysphere.co.jp/en/articles/knowledge-base-content-operations/ Sun, 12 Jul 2026 15:00:00 GMT RAG accuracy is determined not only by the model but also by how quickly document version, expiration, source, permissions, and deletion are reflected in the index; we consider operational design from knowledge benchmarks. Prompt-injection defense starts with permission boundaries, not instruction detection https://catalysphere.co.jp/en/articles/prompt-injection-enterprise-controls/ https://catalysphere.co.jp/en/articles/prompt-injection-enterprise-controls/ Sun, 12 Jul 2026 15:00:00 GMT Filtering suspicious text alone cannot protect AI that reads external data and executes tools; we organize multi-layered defense through attack evaluation, structured input, and official classification. Answering questions about a PDF does not prove document understanding https://catalysphere.co.jp/en/articles/multimodal-document-accuracy/ https://catalysphere.co.jp/en/articles/multimodal-document-accuracy/ Sun, 12 Jul 2026 15:00:00 GMT Being able to extract text via OCR is distinct from correctly linking layout, graphs, tables, and body text; we organize the unit of business verification based on research in document question answering. Using the strongest model for every request is not always the best investment https://catalysphere.co.jp/en/articles/small-model-routing-economics/ https://catalysphere.co.jp/en/articles/small-model-routing-economics/ Sun, 12 Jul 2026 15:00:00 GMT Designing a system that sends all requests to a high-performance model, or one that fixes them to a low-cost model, is not necessarily optimal; drawing from cascade and router research, we consider routing under quality constraints. AI-agent failures persist as state across long workflows https://catalysphere.co.jp/en/articles/ai-agent-workflow-reliability/ https://catalysphere.co.jp/en/articles/ai-agent-workflow-reliability/ Sun, 12 Jul 2026 15:00:00 GMT Even with high accuracy in a single response, errors accumulate in tasks involving continuous planning, search, tool execution, and state updates; we consider the boundaries of reliable automation based on real-world environment benchmarks. Top public benchmark rankings do not determine production quality https://catalysphere.co.jp/en/articles/llm-evaluation-production/ https://catalysphere.co.jp/en/articles/llm-evaluation-production/ Sun, 12 Jul 2026 15:00:00 GMT Rankings from a single benchmark cannot determine production quality, stability, cost, or safety; we must redesign evaluation design by integrating multi-faceted evaluation, human-LLM scoring, and prompt sensitivity research. To what extent can RAG improve the factual accuracy of generative AI? https://catalysphere.co.jp/en/articles/rag-grounding-reliability/ https://catalysphere.co.jp/en/articles/rag-grounding-reliability/ Sun, 12 Jul 2026 15:00:00 GMT Providing evidence via search does not necessarily eliminate errors; this article outlines methods for verifying RAG in business operations by distinguishing between search omissions, inconsistencies with evidence, and oversights in long contexts. The power and water footprint of generative AI cannot be reduced to a per-question figure https://catalysphere.co.jp/en/articles/generative-ai-energy-water/ https://catalysphere.co.jp/en/articles/generative-ai-energy-water/ Sun, 12 Jul 2026 15:00:00 GMT Define measurement boundaries for power and water that include usage, processing volume, model, region, time, and equipment, rather than relying on the impression value of a single prompt, and reflect these in procurement and operations. AI vendor selection must test the ability to resume operations after exit https://catalysphere.co.jp/en/articles/ai-vendor-evaluation-exit/ https://catalysphere.co.jp/en/articles/ai-vendor-evaluation-exit/ Sun, 12 Jul 2026 15:00:00 GMT We organize vendor evaluations that test data, model dependency, contracts, portability, and stoppage/alternatives before implementation, rather than selecting based solely on accuracy demos. Model updates require defining what must be re-proven before the change https://catalysphere.co.jp/en/articles/model-update-change-control/ https://catalysphere.co.jp/en/articles/model-update-change-control/ Sun, 12 Jul 2026 15:00:00 GMT This section outlines a method for managing model, data, search infrastructure, and guardrail updates by linking change scope, comparative evidence, approval, monitoring, and rollback. AI incident response does not end with system recovery https://catalysphere.co.jp/en/articles/ai-incident-response/ https://catalysphere.co.jp/en/articles/ai-incident-response/ Sun, 12 Jul 2026 15:00:00 GMT We outline an incident design that connects damage assessment, causality, evidence of versions/data/decisions, external reporting, and recurrence prevention, rather than ending with model suspension alone. AI fairness cannot be audited with average accuracy and a single score https://catalysphere.co.jp/en/articles/algorithmic-bias-fairness-audit/ https://catalysphere.co.jp/en/articles/algorithmic-bias-fairness-audit/ Sun, 12 Jul 2026 15:00:00 GMT We organize methods for auditing fairness without reducing it to a single score, by examining the usage context, harm, types of errors, intersecting populations, and post-deployment changes. Copyright control for generative AI starts with the production process, not the final output https://catalysphere.co.jp/en/articles/generative-ai-copyright-provenance/ https://catalysphere.co.jp/en/articles/generative-ai-copyright-provenance/ Sun, 12 Jul 2026 15:00:00 GMT We organize methods for separately viewing training, generation, and publication, and for making rights confirmation and provenance recording a single operational audit trail, drawing from primary sources in Japan, the EU, and C2PA. Will generative AI change employment, or will it restructure work first? https://catalysphere.co.jp/en/articles/generative-ai-labor-market-effects/ https://catalysphere.co.jp/en/articles/generative-ai-labor-market-effects/ Sat, 11 Jul 2026 15:00:00 GMT We analyze where changes emerge by dividing the impact of generative AI into task time, wages and working hours of existing employees, freelance demand, and youth hiring. When does AI coding accelerate development, and when does it slow it down? https://catalysphere.co.jp/en/articles/ai-coding-productivity-context/ https://catalysphere.co.jp/en/articles/ai-coding-productivity-context/ Sat, 11 Jul 2026 15:00:00 GMT Average effects of AI coding are not enough for investment decisions. Research from laboratories, enterprises, and mature open-source projects reveals the work, personnel, and measurement conditions that determine whether development speeds up. Does generative AI enhance creativity while homogenizing ideas? https://catalysphere.co.jp/en/articles/generative-ai-creativity-diversity/ https://catalysphere.co.jp/en/articles/generative-ai-creativity-diversity/ Sat, 11 Jul 2026 15:00:00 GMT Even if AI assistance improves individual proposals, if the organization's overall proposals become similar, the exploration range narrows. Based on creativity research, we consider methods to manage quality and collective diversity separately. Does generative AI deepen learning, or does it merely raise performance during support? https://catalysphere.co.jp/en/articles/generative-ai-learning-design/ https://catalysphere.co.jp/en/articles/generative-ai-learning-design/ Sat, 11 Jul 2026 15:00:00 GMT Solving a problem with generative AI is not the same as solving it independently. Drawing on meta-analyses and educational experiments, we examine learning designs that separate supported performance, independent reproduction, and long-term retention. Human–AI collaboration can beat humans alone but does not necessarily beat the best solo performer https://catalysphere.co.jp/en/articles/human-ai-synergy-meta-analysis/ https://catalysphere.co.jp/en/articles/human-ai-synergy-meta-analysis/ Sat, 11 Jul 2026 15:00:00 GMT A meta-analysis of 106 experiments and management studies helps distinguish extension—outperforming humans alone—from true synergy, where collaboration exceeds the best human or AI performance, and identifies when collaboration is viable. Does Generative AI Advice Improve Expert Judgment? https://catalysphere.co.jp/en/articles/llm-professional-decision-quality/ https://catalysphere.co.jp/en/articles/llm-professional-decision-quality/ Sat, 11 Jul 2026 15:00:00 GMT We synthesize results from randomized trials and clinical studies involving physicians to clarify the conditions under which generative AI advice improves judgment, and why mere access does not yield effects. Does Big Data Really Improve Decision-Making? https://catalysphere.co.jp/en/articles/big-data-decision-quality/ https://catalysphere.co.jp/en/articles/big-data-decision-quality/ Sat, 11 Jul 2026 15:00:00 GMT Data volume and decision quality are not the same. We examine the conditions for a usable data foundation, considering metric meaning, causality versus correlation, update frequency, and on-the-ground context. What Is Lost in Organizations That Rely Too Heavily on AI https://catalysphere.co.jp/en/articles/ai-reliance-organizational-risk/ https://catalysphere.co.jp/en/articles/ai-reliance-organizational-risk/ Sat, 11 Jul 2026 15:00:00 GMT Proper support helps; flawed support amplifies errors. Drawing on research into automation bias and critical thinking, we consider how to build organizations that can question, stop, and revert. What Organizations Using AI Are Standardizing https://catalysphere.co.jp/en/articles/ai-adoption-operating-model/ https://catalysphere.co.jp/en/articles/ai-adoption-operating-model/ Sat, 11 Jul 2026 15:00:00 GMT Distributing tools alone does not build organizational capability. Drawing on field research, this article outlines how to redesign work, learning, expertise, and control. The ROI of Generative AI Is Determined by Pre-Implementation Design https://catalysphere.co.jp/en/articles/generative-ai-roi-design/ https://catalysphere.co.jp/en/articles/generative-ai-roi-design/ Sat, 11 Jul 2026 15:00:00 GMT Do not equate productivity gains demonstrated by research directly with business value. We propose designing ROI by linking time, quality, accuracy, and usage costs at the business process level.