Thesis - Human-in-the-Loop Calibrated Agent Harness

## Join our Team About this opportunity AI agents can now create plans, code, and other work much faster than people can review it. When output grows faster than a human can understand and verify it, silent assumptions can accumulate and alignment can be lost. This thesis explores how to keep people actively engaged and empowered while working with capable AI agents—not by maximizing automation or interruption, but by calibrating the level of human involvement to the task. You will investigate and prototype an open, extensible agent harness that supports richer and more natural collaboration between a human and an AI agent. The goal is to move beyond the current, coarse pattern of either letting the agent continue or stopping to ask a question. Instead, the harness should help the agent and human use a shared vocabulary for lightweight clarifications, confidence, assumptions, and decisions, while keeping the human in control of important judgments. What you will do This thesis will combine research, interaction design, software prototyping, and evaluation. You will have freedom to define the project's focus within the overall goal. Areas of exploration could include, but are not limited to: • Designing clarification mechanisms between “say nothing” and “stop and ask a full question,” such as inline assumption flags, confidence indicators, deferred or pinned questions, and accepting a result with a caveat. • Developing an advisor component that monitors the interaction and helps decide when clarification or human attention is valuable. • Exploring how the advisor can support both the main AI agent and the human—for example, by summarizing changes, highlighting decisions, and identifying the parts of an output that mostly need a review. • Investigating how a harness can choose between incremental checkpoints and more autonomous execution based on the cost of verifying the result. • Implementing a prototype in an extensible agent harness, potentially using the PI agent harness and an external model provider. • Evaluating whether a richer, more routinely used interaction vocabulary improves alignment, reduces rework, and helps people maintain understanding and control. The skills you bring • Knowledge in Computer Science, Human-Computer Interaction, Software Engineering, Artificial Intelligence, or a related field. • Experience in at least one programming language and an interest in building and evaluating prototypes. • Interest in large language models, AI agents, human-in-the-loop systems, or interaction design. • Eagerness to explore an interdisciplinary research problem spanning software, AI, and human factors. • Ability to formulate research questions, conduct structured evaluations, and communicate findings clearly. Bonus: experience with Python, TypeScript, LLM APIs, agent frameworks, or user studies. Why join Ericsson?At Ericsson, you´ll have an outstanding opportunity. The chance to use your skills and imagination to push the boundaries of what´s possible. To build solutions never seen before to some of the world’s toughest problems. You´ll be challenged, but you won’t be alone. You´ll be joining a team of diverse innovators, all driven to go beyond the status quo to craft what comes next. What happens once you apply?Click Here to find all you need to know about what our typical hiring process looks like.Encouraging a diverse and inclusive organization is core to our values at Ericsson, that's why we champion it in everything we do. We truly believe that by collaborating with people with different experiences we drive innovation, which is essential for our future growth. We encourage people from all backgrounds to apply and realize their full potential as part of our Ericsson team. Ericsson is proud to be an Equal Opportunity Employer. learn more. Primary country and city: Sweden (SE) || Karlskrona Req ID: 791470