HCOMP 2026 — Human-AI Complementarity & Alignment

HCOMP 2026, the ACM Conference on Human-AI Complementarity and Alignment, runs September 27-30, 2026 at the Virginia Tech Institute for Advanced Computing near Washington, DC. This edition marks a notable repositioning: the long-running Human Computation and Crowdsourcing series has rebranded around human-AI complementarity and alignment, reflecting how the field has moved from studying how people help build AI systems to studying how people and increasingly capable AI work together as complementary partners. The conference is co-located and tightly integrated with the 2026 ACM Collective Intelligence Conference under a shared theme, Connections, encouraging work that bridges disciplines and links humans, animals and machines. For anyone tracking the practical side of the alignment conversation - reliance and trust calibration, scalable human oversight, detecting and repairing misalignment, and dividing labour between people and models across the AI lifecycle - this is one of the few venues treating those questions as an empirical science rather than a manifesto. The program spans human-AI collaboration and co-adaptation, hybrid workflows, human-AI decision-making, human feedback and preference learning (the RLHF lineage), data annotation and quality assurance, bias and responsible-data practice, and governance and accountability in deployed systems.

The disciplinary mix is unusually broad: machine learning and NLP (including LLMs and generative AI) sit alongside HCI design, social computing, computational social science, economics, digital humanities, policy and ethics, so attendees range from ML researchers to qualitative HCI scholars and industry practitioners. Two submission formats structure the program. Full papers (up to 6,000 words) are archival, double-blind, reviewed by three external reviewers under a coordinating PC member, and published in the ACM Digital Library with DOIs; Talks (up to 1,500 words) receive lighter review and remain non-archival, letting social scientists present work bound for journals elsewhere. The conference recognises best paper, best talk and best student work in each track. Co-location with Collective Intelligence roughly doubles the intellectual surface area: CI sessions cover crowds and swarms, collective emotion and polarization, misinformation and belief formation, forecasting and collective decision-making, citizen science and discussion-moderation systems - all relevant to anyone designing human-in-the-loop or multi-agent setups.

Practically, in-person attendance is required for presenters, ACM moved fully to Open Access in 2026, and the venue places attendees a short drive from Washington, DC for those combining the trip with policy or agency visits. HCOMP suits PhD students and faculty in human-centered AI, ML researchers working on human feedback and evaluation, HCI and crowdsourcing specialists, and industry teams building products where humans and models share decisions. If your work concerns how people and AI succeed - or fail - together, HCOMP 2026 is the field's tightest gathering of that community, and one increasingly central as alignment moves from theory into everyday deployment.