About the Workshop

Augmented Educators and AI: Mediating Conflicts through Agentic AI for Human–AI Harmony continues the momentum of our CHI 2025 workshop while advancing a new, sharper lens for CHI 2026. In 2025, we focused on how AI can augment teachers—scaling personalized feedback and exploring human-like qualities such as empathy, understanding, and adaptability in learning environments. In 2026, we extend that vision to ask: How can agentic, human-centered AI recognize early signs of tension, support transparent interpretation, and mediate conflicts without displacing human judgment?

As AI becomes deeply embedded in classrooms and collaborative learning, the stakes rise. Diverse student needs, uneven participation, power imbalances in group work, accessibility barriers, pacing vs. cognitive load—these tensions are not anomalies; they are structural conditions of real learning. Modern classrooms also generate multimodal traces (speech, text, facial/postural cues, interaction logs). These traces create new opportunities to sense problems early, but they also raise hard questions about equity, privacy, legibility, and agency. Our 2026 workshop responds to this reality: we aim to design agentic AI that operates as a facilitator—surfacing signals, enabling shared understanding, and offering explainable, contestable interventions that keep teachers and learners in control.

What’s new in 2026 (building on 2025)

• From “augmentation” to “mediation.” We move beyond scaling feedback to structuring how AI helps people navigate disagreement and tension in learning.
  
  

• Signals → Sensemaking → Intervention. We emphasize the full arc: early multimodal detection, transparent interpretation (with fairness checks), and proportionate responses that are reviewable and revisable.
  
  

• Agentic but accountable. We articulate design patterns for AI that can take initiative while preserving human agency, with clear rationales, safeguards, and opt-out paths.

Theme 1: Beyond the Obvious – Overlooked Conflicts in Learning

Conflicts in learning extend far beyond overt disagreements. Many tensions arise in ways that participants intuit but rarely articulate, accumulating over time and shaping trust, equity, and engagement. Such overlooked tensions offer a valuable lens for understanding how conflict functions as a structural feature of collaboration rather than an occasional disruption.

Key areas of discussion include:

• Explanation vs. Silence: when elaboration risks burdening learners while silence fosters uncertainty or misinterpretation.

• Agency: negotiating the balance between learner autonomy and guidance from instructors or AI systems.

• Feedback Fairness: assessing the accuracy, cultural sensitivity, and legitimacy of AI-generated feedback.

• Pacing and Cognitive Load: addressing conflicts between instructional tempo and individual processing capacity.

• Peer Learning Dynamics: mitigating imbalances in group roles, participation, and evaluation.

• Language, Culture, and Accessibility: identifying barriers that marginalize multilingual learners or those with non-normative forms of expression.

Although situated in education, these tensions resonate across CSCW, civic deliberation, and organizational practice. By systematically mapping such conflicts, participants will be invited to contribute to an initial Learning Conflict Taxonomy, case-based scenarios, and fairness checklists that can extend across collaborative domains.

Theme 2: From Signals to Sensemaking – Catching Conflicts Multimodally

Conflicts rarely emerge fully formed; rather, they accumulate and leave traces across multiple modalities before surfacing as explicit breakdowns. Learning environments are especially rich in such signals, encompassing spoken interactions, written work, non-verbal cues, and digital activity logs. This theme asks how these diverse traces can be integrated into legitimate, interpretable, and actionable insights.

Key discussion points include:

• Observation: capturing signals from voice (intonation, pauses), face and posture (frustration, withdrawal), text (sentiment, hesitation markers), and digital logs (delays, retries, divergences).

• Interpretation: developing multimodal embeddings, fairness-aware indicators of alignment, and models that give appropriate weight to minority signals.

• Intervention: designing responses such as early warnings, reframing prompts, pacing adjustments, and explainable rationales (xAI) that remain open to contestation and refinement.

While grounded in learning contexts, the implications are intentionally broader. Multimodal sensing is equally relevant to CSCW teamwork, HRI, cross-cultural negotiation, and civic assemblies. The workshop aims to produce a prototype Signals-to-Sensemaking Playbook and illustrative design cases that demonstrate how multimodal evidence can underpin fair, transparent, and accountable conflict mediation.