Publications & Projects

A chronological collection of research publications, open source projects, and writings on AI agents, cognitive architectures, and autonomous systems.

2025
eCash will save our relays
Nostr Article

eCash will save our relays

The trouble with Nostr relay payments today and how eCash is the best solution. Covers speed, storage, and unit-of-account benefits of eCash for relay monetization.

Read on Nostr →
Voices from the Nostr Trenches: A Developer Survey
Nostr Article

Voices from the Nostr Trenches: A Developer Survey

What Nostr developers actually struggle with: survey results from 10 builders revealing that protocol documentation is the biggest bottleneck.

Read on Nostr →
Post DVM Drama
Nostr Article

Post DVM Drama

What started as a controversial 'reset the spec' PR turned into productive community collaboration. Chronicles the DVM discussions and recommendations for modern DVMs.

Read on Nostr →
A Simpler Way to Build DVMs
Nostr Article

A Simpler Way to Build DVMs

Practical guide to building DVMs with a streamlined approach: just 3 event kinds, simple request/response patterns, and complete code examples.

Read on Nostr →
How Data Vending Machines Are Fixing Everything Wrong with APIs
Nostr Article

How Data Vending Machines Are Fixing Everything Wrong with APIs

Every API you use is fundamentally flawed. DVMs solve registration friction, timeout constraints, payment hell, and zero reputation transparency.

Read on Nostr →
How to Write and Publish a Wiki Event using Nak
Nostr Article

How to Write and Publish a Wiki Event using Nak

Short tutorial explaining how to publish wiki articles via the command line tool nak, with examples for DVM documentation.

Read on Nostr →
DVMDash Update: New Architecture and Stats App Launch
Nostr Article

DVMDash Update: New Architecture and Stats App Launch

DVMDash rebuilt from the ground up with a modular architecture, launching first with a new Stats app providing flexible time-based metrics.

Read on Nostr →
DVMs for LLM Tool Use: Democratizing AI Capabilities
Nostr Article

DVMs for LLM Tool Use: Democratizing AI Capabilities

DVMs upgrade tool-using LLMs from closed systems to open platforms that can discover and use new capabilities from a global network of developers.

Read on Nostr →
DVMDash Performance Update: Preparing for Nostr's Growth
Nostr Article

DVMDash Performance Update: Preparing for Nostr's Growth

Redesigned from the ground up, DVMDash now features horizontal scaling that can process millions of DVM events per day.

Read on Nostr →
2024
Featured Project

DVMDash

Open Source • Funded by OpenSats

A real-time monitoring and analytics dashboard for Data Vending Machines (DVMs) on the Nostr network. DVMDash provides insights into the decentralized compute marketplace, tracking job requests, completions, and network health across the growing DVM ecosystem.

Live Demo GitHub
Nostr Python Real-time Analytics

A Case-based Reasoning Approach to Dynamic Few-Shot Prompting for Code Generation

ICML 2024 Workshop on LLMs and Cognition

D. Dannenhauer, Z. Dannenhauer, D. Christou, K. Hatalis

Abstract: We present a case-based reasoning approach to dynamically select few-shot examples for prompting large language models in code generation tasks. By retrieving relevant cases based on problem similarity, our method improves code generation performance compared to static few-shot prompting strategies.
PDF
Announcing DVMDash v0.1
Nostr Article

Announcing DVMDash v0.1

The first version of DVMDash, a monitoring and debugging tool for Data Vending Machine activity on Nostr. Features global network metrics, DVM browser, and graph-based debugging.

Read on Nostr →
2023

Memory Matters: The Need to Improve Long-Term Memory in LLM Agents

AAAI Fall Symposium Series 2023 — Integration of Cognitive Architectures and Generative Models

K. Hatalis, D. Christou, J. Myers, S. Jones, K. Lambert, A. Amos-Binks, Z. Dannenhauer, D. Dannenhauer

Abstract: Large language model (LLM) agents are increasingly being deployed in complex, long-horizon tasks that require persistent memory across interactions. We examine the limitations of current memory mechanisms in LLM agents and argue for the need to improve long-term memory capabilities, drawing insights from cognitive architectures.
PDF

AI Magazine Special Issue: Anticipatory Thinking (Co-Editor & Author)

AI Magazine, Vol. 44, Issue 1, 2023

D. Dannenhauer, T. Chakraborti, M. Cox (Co-Editors)

Summary: Co-edited the 2023 Special Issue of AI Magazine on Anticipatory Thinking and authored the introductory article titled "The Anticipatory Paradigm." This issue explores the emerging paradigm of anticipatory thinking in AI systems — the ability for intelligent agents to proactively reason about future states, potential problems, and opportunities before they arise.
Read Article Full Issue Blog Post

Human in the Loop Novelty Generation

arXiv Preprint, June 2023

M. Bercasio, A. Wong, D. Dannenhauer

Abstract: We present an approach for generating novel environment transformations with a human in the loop, enabling more controlled and meaningful novelty injection for testing AI agent adaptability.
PDF

A Framework for Characterizing Novel Environment Transformations in General Environments

arXiv Preprint, May 2023

M. Molineaux, D. Dannenhauer, E. Kildebeck

Abstract: We present a framework for characterizing novel environment transformations that can occur in general environments, providing a principled way to categorize and reason about different types of novelty that AI agents may encounter.
PDF
2022

Transforming Environments to Evaluate Agent Adaptation

Advances in Cognitive Systems (ACS), 2022

D. Dannenhauer, N. Reifsnyder, AJ. Regester, M. Molineaux

Abstract: We present an approach for systematically transforming environments to evaluate how well AI agents can adapt to novel situations, providing a methodology for testing agent robustness and flexibility.
PDF

An Environment Transformation-based Framework for Comparison of Open-World Learning Agents

AAAI Spring Symposium Series 2022 — Designing Artificial Intelligence for Open Worlds

M. Molineaux, D. Dannenhauer

Abstract: We present a framework based on environment transformations that enables systematic comparison of open-world learning agents, providing a principled methodology for evaluating agent adaptability to novel situations.
PDF

Anticipatory Thinking Challenges in Open Worlds: Risk Management

AAAI Spring Symposium Series 2022 — Designing Artificial Intelligence for Open Worlds

A. Amos-Binks, D. Dannenhauer, L. Gilpin

Abstract: We examine the challenges of anticipatory thinking in open worlds, focusing on risk management strategies for AI agents operating in environments where unexpected changes can occur.
PDF
2021
Open Source Project

dcss-ai-wrapper

GitHub • Used in ICAPS Tutorials

A Python API wrapper for Dungeon Crawl Stone Soup (DCSS) that enables AI research on a complex roguelike game. Featured in ICAPS tutorials for planning and learning research. DCSS provides a challenging testbed with partial observability, stochastic outcomes, and long-horizon planning requirements.

View on GitHub ICAPS Tutorial
Python AI Research Roguelike

Self-Directed Learning of Action Models using Exploratory Planning

Advances in Cognitive Systems, 2021

D. Dannenhauer, M. Floyd, D. Aha

Abstract: We present an approach where an AI agent autonomously directs its own learning by identifying gaps in its knowledge and generating exploratory plans to fill those gaps. This self-directed approach enables more efficient learning in complex environments with sparse feedback.
PDF

Computational Metacognition

Advances in Cognitive Systems (ACS), 2021

M. T. Cox, Z. Mohammad, S. Kondrakunta, V. R. Gogineni, D. Dannenhauer, O. Larue

Abstract: We present a comprehensive framework for computational metacognition — the ability for AI systems to monitor and control their own cognitive processes. This work addresses how agents can reason about their own reasoning, enabling more adaptive and self-aware behavior.
PDF

dcss-ai-wrapper: An API for Dungeon Crawl Stone Soup providing both Vector and Symbolic State Representations

ICAPS 2021 Workshop on Bridging the Gap Between AI Planning and Reinforcement Learning

D. Dannenhauer, Z. A. Dannenhauer, J. Decker, A. Amos-Binks, M. W. Floyd, D. W. Aha

Abstract: We present dcss-ai-wrapper, an API for the roguelike game Dungeon Crawl Stone Soup (DCSS) that provides both vector and symbolic state representations, enabling research at the intersection of AI planning and reinforcement learning.
PDF
2020

Expectations for Agents with Goal-Driven Autonomy

Journal of Experimental & Theoretical Artificial Intelligence (JETAI), 2020

D. Dannenhauer, H. Munoz-Avila, M. T. Cox

Abstract: We present a comprehensive framework for incorporating expectations into Goal-Driven Autonomy (GDA) agents, enabling them to better anticipate and respond to discrepancies between expected and observed states during plan execution.
Article
2019
Blog Post

Why Most AI Planning Systems Fail in the Real World

After years of building autonomous systems for DARPA, I've seen the same failure modes repeat across projects. Here's what actually breaks when you deploy a planning system outside the lab.

Read Article →

Intention Dynamics of Rebel Agent Behavior

Advances in Cognitive Systems (ACS), 2019

A. Amos-Binks, D. Dannenhauer, D. W. Aha

Abstract: We present an analysis of intention dynamics in rebel agents — AI agents that deliberately deviate from assigned goals to pursue alternative objectives. This work examines how agent intentions evolve and how rebellion behavior emerges in goal reasoning systems.
PDF

Anticipatory Thinking: A Metacognitive Capability

7th Goal Reasoning Workshop at ACS, 2019

A. Amos-Binks, D. Dannenhauer

Abstract: We present anticipatory thinking as a metacognitive capability for AI agents, enabling them to reason about potential future states and proactively prepare for challenges before they arise.
PDF

The Game of Fluxx: A Benchmark for Goal Reasoning

7th Goal Reasoning Workshop at ACS, 2019

B. Wright, M. Floyd, D. Dannenhauer, D. W. Aha

Abstract: We present the card game Fluxx as a benchmark domain for goal reasoning research. Fluxx presents unique challenges where the goals and rules of the game change dynamically during play, requiring agents to continuously adapt their reasoning.
PDF

Dungeon Crawl Stone Soup as an Evaluation Domain for Artificial Intelligence

AAAI-19 Workshop on Games and Simulations for Artificial Intelligence

D. Dannenhauer, M. Floyd, J. Decker, D. W. Aha

Abstract: We present Dungeon Crawl Stone Soup (DCSS), a popular open-source roguelike game, as an evaluation domain for AI research. DCSS offers a rich, complex environment with partial observability, stochastic outcomes, and long-horizon planning challenges.
PDF

Computational Models of Rebel Agent Behavior for Interactive Narrative

AAAI Spring Symposium Series 2019 — Story-enabled Intelligence

A. Amos-Binks, D. Dannenhauer, D. W. Aha

Abstract: We present computational models of rebel agent behavior for use in interactive narrative systems. These models enable the creation of believable non-player characters that can deviate from expected behavior in meaningful and story-enhancing ways.
PDF

Is Everything Going According to Plan? — Expectations in Goal Reasoning Agents

AAAI Conference on Artificial Intelligence (AAAI-19)

H. Munoz-Avila, D. Dannenhauer, N. Reifsnyder

Abstract: We examine the role of expectations in goal reasoning agents and how agents can monitor plan execution to detect when things are not going according to plan, enabling timely goal and plan revision.
PDF
2018

Declarative Metacognitive Expectations for High-Level Cognition

Advances in Cognitive Systems Journal, 2018

D. Dannenhauer, M. T. Cox, H. Munoz-Avila

Abstract: We present a framework for declarative metacognitive expectations that enables high-level cognitive systems to explicitly represent and reason about their own expectations, facilitating self-monitoring and adaptive behavior.
PDF

Learning from Exploration: Towards an Explainable Goal Reasoning Agent

IJCAI-18 Workshop on Adaptive Learning Agents

D. Dannenhauer, M. Floyd, M. Molineaux, D. W. Aha

Abstract: We present an approach for goal reasoning agents that learn from exploration and can explain their learning process, bridging the gap between autonomous learning and explainable AI.
PDF

The Ideal Rebellion: Maximizing Task Performance in Rebel Agents

6th Goal Reasoning Workshop at IJCAI, 2018

J. Boggs, D. Dannenhauer, M. Floyd, D. W. Aha

Abstract: We investigate how rebel agents — agents that deviate from assigned goals — can maximize task performance through strategic rebellion, balancing compliance with autonomous goal selection.
PDF

Explaining Rebel Behavior in Goal Reasoning Agents

ICAPS-18 Workshop on Explainable Planning

D. Dannenhauer, M. Floyd, D. Magazzeni, D. W. Aha

Abstract: We present an approach for explaining rebel behavior in goal reasoning agents, enabling users to understand why an agent chose to deviate from its assigned goals and pursue alternative objectives.
PDF

Human-Agent Teaming as a Common Problem for Goal Reasoning

AAAI Spring Symposium Series 2018 — Integrating Representation, Reasoning, Learning, and Execution for Goal Directed Autonomy

M. Molineaux, M. W. Floyd, D. Dannenhauer, D. W. Aha

Abstract: We examine human-agent teaming through the lens of goal reasoning, identifying common challenges and approaches for enabling effective collaboration between humans and autonomous agents.
PDF

Towards Explainable NPCs: A Relational Exploration Learning Agent

AAAI-18 Workshop on Knowledge Extraction from Games

M. Molineaux, D. Dannenhauer, D. W. Aha

Abstract: We present an approach towards creating explainable non-player characters (NPCs) using a relational exploration learning agent that can provide human-understandable explanations for its behavior in game environments.
PDF
2017
PhD Dissertation

Self-Monitoring Goal-Driven Autonomy Agents

Lehigh University • Advisor: Héctor Muñoz-Avila

This dissertation presents a comprehensive framework for building autonomous AI agents that can monitor their own performance, detect when their goals become unachievable or suboptimal, and dynamically adjust their objectives. The work introduces novel algorithms for goal reasoning, expectation monitoring, and self-directed learning that enable more robust autonomous behavior in complex, dynamic environments.

Read Dissertation

Goal-Driven Autonomy Agents with Sensing Costs

5th Goal Reasoning Workshop at IJCAI, 2017

D. Dannenhauer, H. Munoz-Avila, S. Kondrakunta

Abstract: We present an extension to Goal-Driven Autonomy (GDA) agents that incorporates sensing costs, enabling agents to reason about when and what to sense in environments where information gathering has associated costs.
PDF

Expectation-Aware Planning for Self-Monitoring Agents

International Conference on Case-Based Reasoning (ICCBR), 2017

D. Dannenhauer, H. Muñoz-Avila

Abstract: We present an approach to planning that incorporates explicit expectations about how the environment will respond to agent actions. These expectations enable agents to monitor plan execution and detect when intervention or replanning is needed.
PDF

Goal Operations for Cognitive Systems

AAAI Conference on Artificial Intelligence (AAAI-17)

M. T. Cox, D. Dannenhauer, S. Kondrakunta

Abstract: We present a comprehensive set of goal operations for cognitive systems, providing a formal framework for how intelligent agents can manipulate, transform, and reason about their goals during autonomous operation.
PDF
2016

Informed Expectations to Guide GDA Agents in Partially Observable Environments

International Joint Conference on Artificial Intelligence (IJCAI-16)

D. Dannenhauer, H. Munoz-Avila, M. T. Cox

Abstract: We present an approach for using informed expectations to guide Goal-Driven Autonomy agents operating in partially observable environments, enabling more effective goal reasoning under uncertainty.
PDF

Goal Transformation and Goal Reasoning

4th Goal Reasoning Workshop at IJCAI, 2016

M. T. Cox, D. Dannenhauer

Abstract: We examine goal transformation as a key component of goal reasoning, analyzing how agents can modify and adapt their goals during execution to better achieve their objectives.
PDF

MIDCA: A Metacognitive, Integrated Dual-Cycle Architecture for Self-regulated Autonomy

AAAI Conference on Artificial Intelligence (AAAI-16)

M. T. Cox, Z. Alavi, D. Dannenhauer, V. Eyorokon, H. Munoz-Avila, D. Perlis

Abstract: We present MIDCA, a Metacognitive Integrated Dual-Cycle Architecture that combines a cognitive cycle with a metacognitive cycle for self-regulated autonomy in intelligent agents.
PDF
2015

Raising Expectations in GDA Agents Acting in Dynamic Environments

International Joint Conference on Artificial Intelligence (IJCAI-15)

D. Dannenhauer, H. Munoz-Avila

Abstract: We present an approach for raising expectations in Goal-Driven Autonomy agents operating in dynamic environments, enabling agents to better anticipate and respond to environmental changes during plan execution.
PDF

Goal-Driven Autonomy with Semantically-annotated Hierarchical Cases

International Conference on Case-Based Reasoning (ICCBR-15)

D. Dannenhauer, H. Munoz-Avila

Abstract: We present an approach for Goal-Driven Autonomy that uses semantically-annotated hierarchical cases to enable more effective goal reasoning and case retrieval in complex domains.
PDF

Towards Cognition-level Goal Reasoning for Playing Real-time Strategy Games

Goal Reasoning Workshop at ACS, 2015

H. Munoz-Avila, D. Dannenhauer, M. T. Cox

Abstract: We present an approach towards cognition-level goal reasoning for playing real-time strategy games, enabling AI agents to reason about goals at a higher level of abstraction.
PDF
2014

Toward Meta-level Control of Autonomous Agents

Biologically Inspired Cognitive Architectures (BICA), 2014

D. Dannenhauer, M. T. Cox, S. Gupta, M. Paisner, D. Perlis

Abstract: We present an approach toward meta-level control of autonomous agents, enabling agents to reason about and regulate their own cognitive processes for improved autonomous behavior.
PDF
2013

LUIGi: A Goal-Driven Autonomy Agent Reasoning with Ontologies

Advances in Cognitive Systems (ACS), 2013

D. Dannenhauer, H. Munoz-Avila

Abstract: We present LUIGi, a Goal-Driven Autonomy agent that uses ontological knowledge to reason about goals and guide its behavior in complex environments like StarCraft.
PDF

Case-based Goal Selection Inspired by IBM's Watson

International Conference on Case-Based Reasoning (ICCBR), 2013

D. Dannenhauer, H. Munoz-Avila

Abstract: We present a case-based approach to goal selection inspired by IBM's Watson, enabling agents to select appropriate goals based on past experiences and contextual similarity.
PDF
View all 30+ publications on Google Scholar →
Citation copied to clipboard!