This paper focuses on the synergistic developmental path between the scientificity and narrativity of science popularization. It systematically explains the mechanism for AIGC-driven synergistic optimization of scientificity and narrativity，analyzes the dual functions of AIGC in enhancing scientificity and optimizing communication，as well as the multidimensional issues behind its usage. Subsequently，it proposes corresponding optimization strategies for issues such as the detrimental effects of AI hallucinations on scientific rigor，the technical bottlenecks in generating highly complex scientific content，and the conflicts between instrumental rationality and value rationality. Namely，establishing multiple mechanisms to suppress AI hallucination generation，conduct narrative transformation for highly complex science popularization content，and build a synergy mechanism between instrumental rationality and value rationality.

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 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.

The popularization of frontier scientific achievements plays a pivotal role in enhancing public understanding and engagementwith science. With the widespread application of generative artificial intelligence（GenAI），its potential performance of popularizing frontier scientific achievements has become a critical topic. Unlike the generation of general science popularization content，GenAI encounters distinct challenges when addressing topics related to frontier scientific achievements. Firstly，there is a paucity of sufficient online textual data for GenAI training.Secondly，these topics often entail relatively higher uncertainty，necessitating a more cautious approach to content formulation. This study conducts a comparative analysis of science popularization content on frontier scientific achievements generated by traditional science popularizers and that produced by GenAI models. The results indicate no significant disparities between the two in terms of scientific accuracy and source reliability. However，the content generated by GenAI models tends to exhibit a more positive emotional tone compared to that produced by traditional science popularizers. Based on these results，this study suggests that while the risk of GenAI disseminating misinformation may be limited，it has the potential to reinforce existing stereotypes. Consequently，GenAI can not entirely supplant traditional science popularizers within the current framework of science popularization.

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.

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.

In addressing the discrepancy of scientists’attitudes and behaviors regarding their participation in science popularization，existing studies have explored the social dimension that influences scientists’intentions. However，they have inadequately addressed the epistemic dimension. This paper focuses on three types of practices：AI + science games developed by scientists，“AI + science”summer lab，and the integration of scientific instrumentation and data for science popularization. These examples illustrate scientists’participation in the age of AI for science. These new patterns of science popularization enable the public to engage directly in scientific epistemic practices. Consequently，scientists actively address the diverse epistemic demands of the public. Such interactions benefit scientific research，collaborative engagement of scientists and the public in scientific practice，and provide opportunities for collaboration among scientists from different fields in science popularization. Furthermore，these practices facilitate the transfer of skills and enhance the public’s understanding of the complexities of science. Ultimately，these initiatives yield mutual epistemic and social benefits for both scientists and the public. This paper proposes it as an effective strategy for integrating epistemic and social approaches to enhance scientists’motivation to participate in science popularization，thereby fostering synergy between science popularization and technological innovation.

To strengthen the national capacity for science popularization and further implement the initiative to enhance the scientific literacy of all citizens，the China Association for Science and Technology（CAST），in collaboration with the National Bureau of Statistics，conducted the 14th national sample survey on Chinese civic scientific literacy in 2024. This survey provided critical data for the evaluation of the“Outline of the National Scheme for Scientific Literacy（2021—2035）”during the 14th Five-Year Plan period，as well as for achieving the goal of raising the proportion of citizens with scientific literacy to over 15% by 2025. The results show that the scientific literacy of Chinese citizens has been improving at an accelerating pace，with the proportion reaching 15.37% in 2024. The imbalance in scientific literacy development has been alleviated to some extent，and the population with basic scientific literacy is now substantial. The internet has further solidified its position as the primary channel for citizens to access scientific and technological information，while science and technology museums have demonstrated higher visitation and utilization efficiency among various science popularization infrastructures. Overall，Chinese citizens increasingly value science，adopt rational and pragmatic attitudes，and support innovation，with further improvements in critical thinking and scientific awareness. The development of artificial intelligence（AI）enjoys a strong public foundation，with citizens holding positive attitudes and ample confidence toward AI applications，though some concerns remain about job displacement.

Generative artificial intelligence（AI）has emerged as a pivotal tool in the creation of science popularization literature; however，its capability in this domain remains subject to skepticism when compared to human authors. To evaluate the sciencepopularization writing competence of generative artificial intelligence，a four-dimensional evaluation framework was developed，comprising readability，engagement，scientific accuracy，and dissemination effectiveness. Using“microbiology”as the thematic focus，sciencepopularization works produced by human creators，DeepSeek，ChatGPT，and ERNIE Bot were systematically examined. The results indicate that both the initial-prompt and deep-prompt versions generated by ChatGPT showed no significant difference in overall scores relative to the human-authored version，with the deep-prompt version exhibiting a significantly superior performance in terms of readability. Moreover，the deep-prompt version of DeepSeek outperformed the human version in the dimensions of engagement and dissemination effectiveness. Notably，evaluators correctly distinguished AI-generated works in less than 55% of cases and showed a tendency to attribute high-scoring works to human creators. These experimental findings suggest that generative artificial intelligence holds the potential to substitute for humanscience popularization creators，thereby prompting the proposal of a novel“human-machine collaborative sciencepopularization creation”model. The study further calls upon the academic community to critically examine the phenomenon of the“Hallucination of Human’s Ability”.

The effective communication of scientific information in popularization short videos requires the collaboration of multiple modal discourses. The integration，complementarity，and reinforcement among these multimodal discourses facilitate audiences’understanding of complex scientific knowledge. Based on a multimodal discourse analysis framework，this study explores the narrative strategies of science popularization short videos intending to enhance their effectiveness. This study selects thescience popularization short videos from the Bilibili account“Huazha Hua Xiaolao”as the research object，analyzing the interaction among visual，linguistic，and auditory modal discourses within the videos. Findings indicate that multimodal narratives enhance the narrative expressiveness of science popularization short videos，stimulate emotional expression，and expand cross-media storytelling capabilities. On this basis，this paper proposes the following narrative strategies：scene-based narrative contextual optimization，deep integration of emotions and narrative cues，and cross-platform multimodal storytelling with full-link dissemination. These strategies aim to provide theoretical and practical guidance for the creation and dissemination of science popularization short videos.