Emotion remains
an elusive andinherently unquantifiable dimension of human experience.
AI MYEYESII：Emotional Mimicry
人工智能与我的双眼：情绪拟态
「"Emotional mimicry"（情绪模仿）是心理学和社会学领域的一个重要概念，指的是个体在与他人交往过程中，无意识地模仿和反映对方的情绪表达和情感状态的行为。这种模仿行为可以体现在面部表情、语音的音调、身体姿态和动作等方面。情绪模仿是人类社会交往中的一种基本机制，有助于促进人际间的理解和情感共鸣，増强社会联系和群体的凝聚力。」
--厄休拉·埃斯(Ursula Hess)
「AI MY EYES II：Emotional Mimicry /人工智能与我的双眼：情绪拟态」旨在探索人工智能是如何通过创造“拟像(Simulacrum)”的方式，进而以视觉角度影响，甚至干涉人类对于阅读图像时所成情感的感知。
项目通过数据采集，信息提取，拟像生成这三步骤构成。在数据采集的部分，以游戏摄影的方式，即在游戏环境中捕捉具有情绪色彩的瞬间，包括角色的表情、身体语言及通过环境、色彩、光线传达的情绪氛围，作为保留虚拟世界中情绪的存储切片。在项目前身——AI MY EYES：Interweaving Realms 中，我将这些游戏摄影图像作为数据集交付给到经过训练的人工智能系统GAN（对抗式生成模型）中，再对其生成内容进行筛选后合成。
然而，在本次续作项目——AI MY EYES II：Emotional mimicry中，筛选游戏摄影图像后，交由给CLIP (Contrastive Language–Image Pre-training) 模型进行图像语义理解，这一系统通过文本重现了图像的视觉特征。再通过迭代后的扩散模型算法（Diffusion Model），将分析出对文本以新的“拟像”形态生成，最终借由数字合成的手段将原图像与“拟像”融合，使其从视觉上产生一种和谐的共生状态。利用人工智能对图像的理解（将信息从数据中提取）与再生成（将信息转化为原数据媒介），完成“拟像生成”这一步骤，经过与原图像的合成后，每一幅游戏摄影图像都在其原有基础上获得了视觉情绪上的加深与画面观感上的扩展。
即便人工智能的图像分析模型并不完全理解人类的情绪复杂性，但经过其数据分析后，所反馈的符合普世价值理解的文本，与利用其文本进而生成的“拟象”，依旧能够激发观看者对画面中的情绪共情和自然秩序的联想，这种单纯的媒介转化，其产物却下意识地符合了“无意识地模仿和反映对方的情绪表达和情感状态的行为”这一概念，从而达成了一种新型的“情绪模仿(Emotional Mimicry)”。
「情绪，是自始至终存在于人性之中，且难以量化的一种状态。项目希望通过AI MY EYES系列中对人类情感与机器感知之间关系的探讨，去引发观众对于视觉艺术、技术和情感共鸣的思考。同时探索在这个处于人工智能爆发前夜奇点的当下，人类情感的存在与客观事实下人性的构成，在科技与艺术媒介中具有一种怎样的辩证关系。」
“Emotional mimicry,” as articulated by Ursula Hess, refers to the unconscious tendency of individuals to mirror and reflect the emotional expressions and affective states of others. This process manifests through facial expressions, vocal intonation, posture, and gesture. It constitutes a fundamental mechanism of human social interaction, facilitating empathy, mutual understanding, and collective cohesion.
AI MY EYES II: Emotional Mimicry extends this concept into the domain of artificial intelligence, investigating how machine-generated simulacra can influence—even subtly interfere with—the formation of emotional perception in the act of viewing images. The project situates itself within a broader inquiry into the mediation of affect, asking how emotion may be translated, reconstructed, and re-experienced through computational systems.
The work unfolds through three interrelated stages: data acquisition, information extraction, and simulacral generation. In the initial phase, emotionally charged moments are captured through in-game photography. These images register expressions, bodily gestures, and atmospheres shaped by environment, color, and light, functioning as preserved fragments of affect within virtual worlds. In the project’s precursor, AI MY EYES: Interweaving Realms, these images were used as datasets and processed through a Generative Adversarial Network, where generated outputs were curated and composited.
In this subsequent iteration, selected images are first processed by CLIP, which performs semantic interpretation by translating visual features into textual descriptions. These textual outputs are then fed into an iterative Diffusion Model, producing new visual forms as simulacra. Finally, through digital compositing, the original images and their generated counterparts are merged, establishing a condition of visual coexistence. Within this process, artificial intelligence performs a dual operation: extracting latent information from image data and re-inscribing it into a regenerated visual medium. Each resulting image thus acquires an intensified emotional register and an expanded perceptual field.
Although contemporary image analysis systems do not fully comprehend the complexity of human emotion, the textual interpretations they generate—aligned with broadly shared semantic frameworks—along with the derived simulacra, nonetheless evoke affective resonance and associations with natural order. This act of translation across media, while technically reductive, paradoxically enacts a form of “unconscious mirroring.” In this sense, the project proposes a technologically mediated extension of emotional mimicry, wherein machines participate in the reflection and amplification of affect.
Emotion remains an elusive and inherently unquantifiable dimension of human experience. Through the AI MY EYES series, this project seeks to examine the relational dynamics between human affect and machine perception, inviting reflection on the intersections of visuality, technology, and empathy. At a historical moment approaching the threshold of widespread artificial intelligence integration, it further considers the dialectical interplay between human subjectivity and objective systems, and how this tension reshapes the conditions through which emotion is perceived, constructed, and shared.
2024