As I near the end of my final semester for my master of library and information science (MLIS) program at SJSU, I recently traveled to Seattle for the 2026 ACM SIGIR Conference on Human Information Interaction and Retrieval (CHIIR). Coming from a background in enterprise product marketing and knowledge management, I attended this conference to gain a deeper understanding of information retrieval.

The conference made one thing clear; we are moving away from ranked lists of links and toward agentic systems that prioritize task completion. Below are some key takeaways that I gathered from the conference.

The shift to agentic information retrieval

Traditional search has always placed a high cognitive load on the user. We type a query, get a list of links, and then we must compare sources and synthesize the evidence ourselves. The main theme at CHIIR 2026 was the transition to Agentic AI—systems that can perceive, reason, and act autonomously to solve problems.

In this new paradigm, the unit of interaction is no longer the query; it is task completion. Instead of just document retrieval, we are seeing a shift toward goal-oriented workflows where the AI decomposes a multifaceted request into a hierarchy of simpler tasks.

From tool to teacher

One of the most impactful sessions, From Tool to Teacher, challenged the foundation of current search systems. Most systems assume that users have perfectly formed questions and the experience to distinguish signal from noise—assumptions that often fail when dealing with complex health information or controversial topics.

The research suggests that we should rethink search systems as instructive interfaces. Rather than just providing an answer, a system can act as a coach using a didactic toolkit with three core components.

• Scaffolding: provides temporary support for complex tasks and gradually fades away as the user learns.
• Cognitive apprenticeship: Making expert thinking visible to teach users how to evaluate information themselves.
• Instructional feedback: Providing targeted information to help users improve their search performance.

Human-centered AI and the agency gap

As these agents become more autonomous, the agency gap becomes a critical concern for information professionals. There was a strong argument at the conference that the term human-AI collaboration can be misleading because only people and organizations can be held accountable for information outcomes.

To address this, human-centered AI (HCAI) emphasizes tools that amplify and enhance human performance rather than replacing it. An effective design for these systems should consider the following principles. 

• Predictability and control: Users need compact visual control panels where they can see the impact of their decisions in real-time.
• Recognition over recall: Use buttons, sliders, and pulldowns to facilitate understanding instead of relying purely on natural language prompts.

The challenge of temporal drift

For those of us interested in the technical side of information management, the concept of temporal drift was a recurring warning that surfaced in a workshop on agentic AI for information access. This occurs when stored facts or AI memories become obsolete because the underlying information in the real world has changed. This is a major hurdle for agentic systems that rely on mixed, multi-step pipelines, and it requires us to develop new benchmarks to ensure systems are evaluated on fresh, unseen data.

Final thoughts

Attending CHIIR 2026 underscored that even as tools change, core MLIS skills—evaluating authority, understanding user needs, and managing complex knowledge systems—are more vital than ever. Whether building chatbots or managing digital archives, we now govern how autonomous systems and humans’ access and present information.

For fellow students, I highly recommend applying for iSchool travel grants. These conferences provide a necessary bridge between our coursework and the experiences we will have in the real world.