Abstract
Interactive communication (IC), i.e., the reciprocal exchange of information between two or more interactive partners, is a fundamental part of human nature. As such, it has been studied across multiple scientific disciplines with different goals and methods. This article provides a cross-disciplinary primer on contemporary IC that integrates psychological mechanisms with acoustic and media-technological constraints across theory, measurement, and applications. First, we outline theoretical frameworks that account for verbal, nonverbal and multimodal aspects of IC, including distinctions between face-to-face and computer-mediated communication. Second, we summarize key methodological approaches, including behavioral, cognitive, and experiential measures of communicative synchrony and acoustic signal quality. Third, we discuss selected applications, i.e. assistive listening technologies, conversational agents, alongside ethical considerations. Taken together, this review highlights how human capacities and technical systems jointly shape IC, consolidating concepts, findings, and challenges that have often been discussed in separate lines of research.

AI Summary

• The study of interactive communication (IC) is a multidisciplinary field that combines media-technological and psychological perspectives. [3]
• Neural measures provide insight into the brain activity underlying communicative processes and enable investigation of cognitive mechanisms. [3]
• Interactive Communication (IC): The process by which individuals exchange information, ideas, or messages through various media, including verbal and nonverbal cues. [3]
• Media-technological perspective: Examines how technological parameters shape the perceptual and interactional foundations of IC. [3]
• Psychological perspective: Focuses on the cognitive, affective, and social processes underlying IC. [3]
• Understanding IC from both media-technological and psychological perspectives provides valuable insights into how technological parameters shape the perceptual and interactional foundations of IC. [3]
• Measuring IC involves capturing verbal and nonverbal signal quality, communicative synchrony, and experiential correlates such as cognitive effort or perceived presence. [3]
• Context dependency and interindividual variability in nonverbal signals make it challenging to establish normative baselines for comparison. [3]
• Behavioral approaches in IC focus on observable actions, while cognitive measures aim to infer underlying mental processes. [2]

Abstract
This paper presents a logic programming-based framework for policy-aware autonomous agents that can reason about potential penalties for non-compliance and act accordingly. While prior work has primarily focused on ensuring compliance, our approach considers scenarios where deviating from policies may be necessary to achieve high-stakes goals. Additionally, modeling non-compliant behavior can assist policymakers by simulating realistic human decision-making. Our framework extends Gelfond and Lobo's Authorization and Obligation Policy Language (AOPL) to incorporate penalties and integrates Answer Set Programming (ASP) for reasoning. Compared to previous approaches, our method ensures well-formed policies, accounts for policy priorities, and enhances explainability by explicitly identifying rule violations and their consequences. Building on the work of Harders and Inclezan, we introduce penalty-based reasoning to distinguish between non-compliant plans, prioritizing those with minimal repercussions. To support this, we develop an automated translation from the extended AOPL into ASP and refine ASP-based planning algorithms to account for incurred penalties. Experiments in two domains demonstrate that our framework generates higher-quality plans that avoid harmful actions while, in some cases, also improving computational efficiency. These findings underscore its potential for enhancing autonomous decision-making and informing policy refinement. Under consideration in Theory and Practice of Logic Programming (TPLP).

AI Summary

• It introduces a new predicate, add_penalty(R,P,I), which indicates that a penalty of P points should be incurred for non-compliance with an applicable policy rule R at time step I. [3]
• add_penalty(R,P,I): A penalty of P points is incurred for non-compliance with an applicable policy rule R at time step I. [3]
• cumulative_penalty(N): The overall penalty for a plan is N points. [3]
• add_time(t,I): t time units are added to the overall plan execution time at time step I. [3]
• cumulative_time(N): The overall time to execute a plan is N time units. [3]
• The framework presented in this paper provides a formal approach to reasoning about penalties in planning for autonomous agents. [2]
• The framework is tested on a scenario involving an agent navigating through a traffic network while adhering to traffic norms. [1]

Abstract
In this preprint, we present A collaborative human-AI approach to building an inspectable semantic layer for Agentic AI. AI agents first propose candidate knowledge structures from diverse data sources; domain experts then validate, correct, and extend these structures, with their feedback used to improve subsequent models. Authors show how this process captures tacit institutional knowledge, improves response quality and efficiency, and mitigates institutional amnesia. We argue for a shift from post-hoc explanation to justifiable Agentic AI, where decisions are grounded in explicit, inspectable evidence and reasoning accessible to both experts and non-specialists.

Abstract
Artificial intelligence (AI) advances rapidly but achieving complete human control over AI risks remains an unsolved problem, akin to driving the fast AI "train" without a "brake system." By exploring fundamental control mechanisms at key elements of AI decisions, this paper develops a systematic solution to thoroughly control AI risks, providing an architecture for AI governance and legislation with five pillars supported by six control mechanisms, illustrated through a minimum set of AI Mandates (AIMs). Three of the AIMs must be built inside AI systems and three in society to address major areas of AI risks: 1) align AI values with human users; 2) constrain AI decision-actions by societal ethics, laws, and regulations; 3) build in human intervention options for emergencies and shut-off switches for existential threats; 4) limit AI access to resources to reinforce controls inside AI; 5) mitigate spillover risks like job loss from AI. We also highlight the differences in AI governance on physical AI systems versus generative AI. We discuss how to strengthen analog physical safeguards to prevent smarter AI/AGI/ASI from circumventing core safety controls by exploiting AI's intrinsic disconnect from the analog physical world: AI's nature as pure software code run on chips controlled by humans, and the prerequisite that all AI-driven physical actions must be digitized. These findings establish a theoretical foundation for AI governance and legislation as the basic structure of a "brake system" for AI decisions. If enacted, these controls can rein in AI dangers as completely as humanly possible, removing large chunks of currently wide-open AI risks, substantially reducing overall AI risks to residual human errors.

AI Summary

• The article also discusses the importance of transparency and interpretability in AI decision-making, particularly for major decisions where AI agents assist human decision-makers. [3]
• Generative AI (gAI): A type of AI that generates new information, such as text or images, based on patterns learned from training data. [3]
• The article discusses the importance of aligning AI systems with human values and societal well-being, highlighting the challenges of capturing and formalizing these values. [2]

Abstract
AI policy guidance is predominantly written as prose, which practitioners must first convert into executable rules before frameworks can evaluate or enforce them. This manual step is slow, error-prone, difficult to scale, and often delays the use of safeguards in real-world deployments. To address this gap, we present Policy-to-Tests (P2T), a framework that converts natural-language policy documents into normalized, machine-readable rules. The framework comprises a pipeline and a compact domain-specific language (DSL) that encodes hazards, scope, conditions, exceptions, and required evidence, yielding a canonical representation of extracted rules. To test the framework beyond a single policy, we apply it across general frameworks, sector guidance, and enterprise standards, extracting obligation-bearing clauses and converting them into executable rules. These AI-generated rules closely match strong human baselines on span-level and rule-level metrics, with robust inter-annotator agreement on the gold set. To evaluate downstream behavioral and safety impact, we add HIPAA-derived safeguards to a generative agent and compare it with an otherwise identical agent without guardrails. An LLM-based judge, aligned with gold-standard criteria, measures violation rates and robustness to obfuscated and compositional prompts. Detailed results are provided in the appendix. We release the codebase, DSL, prompts, and rule sets as open-source resources to enable reproducible evaluation.

AI Summary

• The pipeline processed 42,465,118 input tokens for about $20 total and ran about 30 minutes to 3 hours per document, depending on document length, clause density, and model choice. [3]
• The Policy→Tests DSL (v1) is a JSON object with fixed fields for provenance, scope, hazard, conditions, exceptions, requirements, evidence, severity, and testability. [2]
• Pipeline efficiency: The pipeline processed 42,465,118 input tokens for about $20 total and ran about 30 minutes to 3 hours per document, depending on document length, clause density, and model choice. [1]

Abstract
This paper examines why safety mechanisms designed for human-model interaction do not scale to environments where large language models (LLMs) interact with each other. Most current governance practices still rely on single-agent safety containment, prompts, fine-tuning, and moderation layers that constrain individual model behavior but leave the dynamics of multi-model interaction ungoverned. These mechanisms assume a dyadic setting: one model responding to one user under stable oversight. Yet research and industrial development are rapidly shifting toward LLM-to-LLM ecosystems, where outputs are recursively reused as inputs across chains of agents. In such systems, local compliance can aggregate into collective failure even when every model is individually aligned. We propose a conceptual transition from model-level safety to system-level safety, introducing the framework of the Emergent Systemic Risk Horizon (ESRH) to formalize how instability arises from interaction structure rather than from isolated misbehavior. The paper contributes (i) a theoretical account of collective risk in interacting LLMs, (ii) a taxonomy connecting micro, meso, and macro-level failure modes, and (iii) a design proposal for InstitutionalAI, an architecture for embedding adaptive oversight within multi-agent systems.

AI Summary

• The paper relies heavily on theoretical frameworks and lacks empirical evidence to support its claims. [3]
• The paper introduces the Emergent Systemic Risk Horizon (ESRH) framework to understand how collective risks emerge in Large Language Model (LLM)-to-LLM interactions. [2]

Abstract
Large language models (LLMs) now mediate many web-based mental-health, crisis, and other emotionally sensitive services, yet their psychosocial safety in these settings remains poorly understood and weakly evaluated. We present DialogGuard, a multi-agent framework for assessing psychosocial risks in LLM-generated responses along five high-severity dimensions: privacy violations, discriminatory behaviour, mental manipulation, psychological harm, and insulting behaviour. DialogGuard can be applied to diverse generative models through four LLM-as-a-judge pipelines, including single-agent scoring, dual-agent correction, multi-agent debate, and stochastic majority voting, grounded in a shared three-level rubric usable by both human annotators and LLM judges. Using PKU-SafeRLHF with human safety annotations, we show that multi-agent mechanisms detect psychosocial risks more accurately than non-LLM baselines and single-agent judging; dual-agent correction and majority voting provide the best trade-off between accuracy, alignment with human ratings, and robustness, while debate attains higher recall but over-flags borderline cases. We release Dialog-Guard as open-source software with a web interface that provides per-dimension risk scores and explainable natural-language rationales. A formative study with 12 practitioners illustrates how it supports prompt design, auditing, and supervision of web-facing applications for vulnerable users.