The burgeoning of generative artificial intelligence is reshaping the paradigm of scientific knowledge dissemination，while its derivative phenomenon of “knowledge hallucination” profoundly deconstructs the foundational credibility of scientific communication. Drawing upon Latour’s actornetwork theory analytical framework，this paper posits AI knowledge hallucination as a product of systemic distortion or “betrayal” within the translation process. This occurs within a heterogeneous actornetwork due to the divergence of objectives and conflicts of interest among key actors. This phenomenon stems from four structural dysfunctions：Source networks suffer contamination from diachronic biases in data inscription and algorithmic goal alienation；Error-correction networks face institutional feedback deficits and temporal asynchrony conflicts hindering rectification; Translation chains endure professional discourse dimensionality loss and technical violence from cross-contamination within heterogeneous knowledge networks；Accountability networks descend into governance vacuums due to technological black-box obscurity and fragmented responsibility. To address these issues，this paper proposes a four-dimensional path for collaborative governance：“Source Purification” fortifies data foundations through a multi-centre knowledge certification system and blockchain traceability technology；“Process Streamlining” builds a closed-loop error correction network leveraging intelligent monitoring systems and dynamic knowledge repositorie；“Translation Optimisation” ensures knowledge fidelity via context-aware algorithms and human-machine dual verification mechanisms；“Anchoring Accountability” clarifies responsibility boundaries among diverse stakeholders through legal empowerment and transparent algorithmic disclosure. This research deepens understanding of the socio-technical nature of AI knowledge hallucination，advancing the development of a robust，trustworthy “AI+Science Communication” ecosystem centred on human-machine collaboration.

Deep synthesis technologies，exemplified by DeepSeek，are injecting powerful momentum into modern science communication，guiding the public’s understanding of science into a new normal. They are propelling public cognitive paradigms from “tool-assisted” to a novel stage of “cognitive enhancement”. However，this technological leap is not entirely smooth. While presenting numerous opportunities，it also harbors cognitive risks. This inherent tension reveals that while the public benefits from enhanced cognitive efficiency，expanded cognitive horizons，and transformed knowledge transmission paradigms through algorithmic tools like DeepSeek，they simultaneously face risks of cognitive displacement which includes systemic degradation of critical thinking，pathological fixation of cognitive dependencies，and hierarchical rigidification of cognitive disparities due to “cognitive outsourcing”. Faced with the technological empowerment value and potential cognitive risks demonstrated by tools like DeepSeek in AI science popularization，we must pursue a rational dialectical thinking dimension to advance development，constructe a three-dimensional collaborative framework integrating technology，cognition，and ethics. By restructuring critical thinking training mechanisms，refining technical usage standards，and strengthening social equity safeguards，we can foster a new ecosystem for AI science communication characterized by human-machine symbiosis and inclusive sharing，ensure technological capabilities genuinely serve the comprehensive development of the public as DeepSeek and similar deep synthesis service algorithms integrate with modern science communication.

The application of generative artificial intelligence in the field of science writing is burgeoning. While empowering science writing，it also introduces unique challenges，such as scientific inaccuracies of science popularization works，the identity and ability recognition of science writers，and the “limited autonomy” of the audience under the implicit dominance of technology. It is suggested to guide the responsible and sustainable development of generative AI-assisted science writing from several aspects：Adhering to the bottom line of the scientificity of science popularization works；Strengthening the subjectivity of “human” of science writers and emphasizing the organic integration of instrumental rationality and value rationality；Enhancing the scientific and digital literacy of the public and creating a new environment for science writing that suits the current situation.

With the widespread application of generative artificial intelligence in creative domains，AI is increasingly recognized as an autonomous co-creator，making human-AI co-creation a frontier topic in creativity research. Unlike conventional AI-assisted models，genuine human-AI co-creation requires humans to shift from task executors to “directors” and “gatekeepers”，maintaining proactive agency and leveraging their irreplaceable strengths in divergent thinking，contextual judgment，and authenticity control. However，current practices often fail to fully adhere to these critical boundaries，creative human-machine collaboration unleashs individual creative potential and promoting equality in creativity，and led to challenges accompanying the benefits concurrently，such as homogenization effects，weakened interpersonal collaboration，and copyright disputes. To promote sustainable development in human-AI co-creation，it is imperative to reshape public perceptions of creativity，advance evidence-based human-AI collaborative practices，systematically cultivate co-creation literacy，construct an ecosystem for human-human-AI co-creation，and establish ethical and institutional frameworks that balance democratization with incentives.