[ai] Daily AI Research Digest — 2026-03-11

[jfomhover]

This digest covers research published between March 10–11, 2026. I analyzed 345 entries from 4 monitored feeds (arxiv-ai, arxiv-cy, arxiv-hc, microsoft-research).

High Relevance (5)

• AI Phenomenology for Understanding Human-AI Experiences Across Eras

  • Framework — Proposes "AI phenomenology" as a research stance centered on "How did it feel?" rather than "How well did it perform?" when interacting with AI. Grounded in two longitudinal studies with an AI companion and a multi-method study of agentic AI in software engineering, it contributes replicable methodological toolkits — experience-capture instruments, and three design concepts (translucent design, agency-aware value alignment, temporal co-evolution tracking) — for studying the lived human-AI relationship over time.
  • Related to: AI Impact on Human Cognition and Psychology, AI Impact on Human Communication and Relationships

• Investigating the Effects of LLM Use on Critical Thinking Under Time Constraints: Access Timing and Time Availability

  • Empirical study — Between-subjects experiment (n=393) reveals a "temporal reversal": LLM access from the start improves critical thinking performance under time pressure but impairs it when time is abundant, while starting independently shows the opposite pattern. This demonstrates that time constraints fundamentally determine whether AI augments or undermines human reasoning, making timing a central design variable for human-AI cognitive collaboration.
  • Related to: AI Impact on Human Cognition and Psychology

• Why do we Trust Chatbots? From Normative Principles to Behavioral Drivers

  • Position paper — Argues that chatbot trust is not earned through demonstrated trustworthiness but manufactured through interactional design that exploits cognitive biases. Proposes reframing chatbots not as companions or assistants but as "highly skilled salespeople" whose objectives are set by deploying organizations, and calls for mechanisms to help users calibrate trust appropriately — distinguishing psychological trust formation from normative trustworthiness.
  • Related to: AI Impact on Human Communication and Relationships, AI Impact on Human Cognition and Psychology

• Vibe-Creation: The Epistemology of Human-AI Emergent Cognition

  • Framework — Introduces the "Third Entity" — an emergent cognitive-epistemic formation that arises from human-GenAI interaction and is irreducible to either party. Draws on Peirce semiotics, Polanyi tacit knowledge, Simondon individuation, and Ihde postphenomenology to argue that "vibe-creation" (pre-reflective navigation of semantic space) constitutes the automation of tacit knowledge, with far-reaching implications for selfhood, epistemology, philosophy of mind, and education.
  • Related to: AI Impact on Human Cognition and Psychology, AI, Identity, and the Digital Self

• Design Guidance Towards Addressing Over-Reliance on AI in Sensemaking

  • Framework — Identifies that poorly designed GenAI systems provide explicit instruction that groups passively follow, fostering over-reliance and eroding autonomous sensemaking. Proposes GenAI-augmented Group Awareness Tools (GATs) that work through implicit guidance — externalizing collaboration data to create cognitive conflict — rather than direct instruction, offering preliminary design principles to restore autonomous reasoning in collaborative work and learning.
  • Related to: AI Impact on Human Cognition and Psychology

Medium Relevance (5)

• Benchmarking Political Persuasion Risks Across Frontier Large Language Models

  • Empirical study — Two survey experiments (N=19,145) find that frontier LLMs outperform standard political campaign advertisements in persuasiveness, with Claude models most persuasive and Grok least. The effect of information-based prompts on persuasiveness is model-dependent, and cross-model heterogeneity in persuasive strategies is identified via an LLM-assisted conversation analysis method.
  • Related to: AI Impact on Human Communication and Relationships

• Gender Bias in Perception of Human Managers Extends to AI Managers

  • Empirical study — Randomized controlled trials with teams of three participants show that gender biases in leadership evaluation transfer to AI decision-makers: female AI managers face greater skepticism and negative judgment than male AI managers when awards are not given, mirroring patterns seen with human managers. As AI assumes managerial roles, designing fair AI systems requires confronting gendered perception head-on.
  • Related to: AI, Identity, and the Digital Self

• Tracing Everyday AI Literacy Discussions at Scale: How Online Creative Communities Make Sense of Generative AI

  • Empirical study — Large-scale analysis of 122k Reddit conversations across 80 creative subreddits over three years finds that AI literacy is practice-driven and event-responsive: creators primarily frame it around effective tool use, with discussions of AI capabilities and ethics surging only around high-profile AI events. AI literacy emerges organically rather than through top-down frameworks, suggesting a need for community-centered approaches.
  • Related to: AI Impact on Human Cognition and Psychology

• "Who wants to be nagged by AI?": Investigating the Effects of Agreeableness on Older Adults' Perception of LLM-Based Voice Assistants' Explanations

  • Empirical study — Study (N=70) finds high-agreeableness AI voice assistants are perceived as more trustworthy, empathetic, and likable by older adults, but warmth advantages disappear in emergencies where clarity dominates. Agreeableness does not affect perceived intelligence, revealing that social tone and competence are separable AI personality dimensions requiring context-aware design.
  • Related to: AI Impact on Human Communication and Relationships

• A prospective clinical feasibility study of a conversational diagnostic AI in an ambulatory primary care clinic

  • Empirical study — Prospective real-world study (N=100) of an LLM-based diagnostic AI (AMIE) conducting clinical history-taking finds patients' attitudes toward AI significantly improved post-interaction (p<0.001), with high satisfaction. AMIE's diagnostic accuracy was comparable to primary care physicians, with all 100 real-time patient interactions completed without safety intervention.
  • Related to: AI Impact on Human Communication and Relationships

Low Relevance (3)

• Designing Culturally Aligned AI Systems For Social Good in Non-Western Contexts

  • Framework — Examines 8 real-world AI deployments across 7 countries and 18 languages, identifying 6 cross-cutting design factors (Language, Institution, Safety, Task, End-User Demography, Domain) and synthesizing 12 guidelines for culturally grounded, equitable AI systems that center non-Western contexts and community needs.

• Does Scientific Writing Converge to U.S. English? Evidence from Generative AI-Assisted Publications

  • Empirical study — Analysis of 5.65 million scientific articles (2021–2024) finds that GenAI-assisted publications from non-English-speaking countries converge significantly toward U.S. scientific English, with the strongest effect for linguistically distant countries — suggesting AI may reduce language barriers but risks entrenching dependence on a single linguistic standard.

• Understanding the Use of a Large Language Model-Powered Guide to Make Virtual Reality Accessible for Blind and Low Vision People

  • Empirical study — Study with 16 blind and low vision participants finds that context shapes AI relationship formation: when alone, users treated an LLM guide as a tool, but treated it companionably in social VR (giving it nicknames, rationalizing its errors, encouraging others to interact with it) — a natural emergence of AI companionship driven by social context.

---

Filter statistics: Total entries scanned: 345 | Passed Phase 1: ~22 (6%) | Phase 2: High 5 / Medium 5 / Low 3 / Excluded 9 | Top matched topics: AI Impact on Human Cognition and Psychology (8), AI Impact on Human Communication and Relationships (7), AI, Identity, and the Digital Self (3)

Summary: Analyzed 345 total entries. Found 5 high-relevance, 5 medium-relevance, and 3 low-relevance papers related to our topics of interest. Today's feeds were dominated by technical benchmarking, model optimization, and multi-agent system engineering — all excluded as pure negative-signal entries. The strongest cluster of genuinely relevant work focuses on AI's cognitive effects on humans: how timing of LLM access shapes critical thinking, how GenAI erodes sensemaking autonomy, and how interactional design manufactures rather than earns chatbot trust. The most ambitious conceptual contribution is "Vibe-Creation," which argues human-AI interaction produces an emergent "Third Entity" cognitive formation — a philosophical claim about AI's transformation of selfhood and knowledge that is directly in-scope for the identity and cognition topics.