From Perceptual Cues to Screening Judgments: How Users Evaluate AI-Generated Interface Candidates

This dataset supports a sequential empirical study on how users evaluate AI-generated interface candidates after generation. The study examines AI-generated game main-interface candidates as cue-based evaluation objects and investigates how validated perceptual cues are associated with preference selection, post-choice aesthetic appraisal, perceived acceptance, and early-stage screening judgments. The dataset consists of three anonymized supplementary data workbooks. Supplementary Data 1 contains the Stage 3 stimulus-validation dataset and statistical outputs. In this stage, 52 participants independently evaluated 16 AI-generated game main-interface stimuli labeled S1–S16. The stimuli were rated on five perceptual cue dimensions: visual order, visual variation, atmospheric richness, generated-interface credibility, and perceived AI-generated imperfection. The workbook includes stimulus-level validation scores, reliability results, descriptive statistics, and validation checks for the intended visual order × visual variation cue structure. Supplementary Data 2 contains the Stage 4 main user-evaluation dataset and derived analysis results. This stage retained 515 valid responses. Participants selected their preferred interface candidate from the 16 AI-generated stimuli and then evaluated only the selected stimulus using post-choice aesthetic dimensions based on simplicity, diversity, colorfulness, and craftsmanship. The workbook includes anonymized participant-level responses, neutral S1–S16 stimulus-ID coding, stimulus-type mapping, selected-stimulus cue mapping, preference-selection distributions, reliability and construct-adequacy results, cue–rating correlations, composite cue–appraisal associations, perceived-acceptance analyses, and exploratory user-level comparison results. The S1–S16 labels are neutral stimulus identifiers and do not correspond to MBTI types, cognitive-style groups, or personality categories. Supplementary Data 3 contains the Stage 5 follow-up mechanism-check dataset and mixed-effects model outputs. Stage 5 used a voluntary follow-up subsample of 200 participants from the Stage 4 respondent pool. Each participant evaluated eight representative AI-generated interface candidates, resulting in 1,600 stimulus-level observations in long format. The dataset includes item-level responses and scale scores for processing ease, visual cognitive effort, interface trust, continued exploration intention, and development-worthiness. It also includes stimulus metadata, validated cue scores mapped from Stage 3, derived cue indices, reliability and descriptive statistics, correlation checks, mixed-effects model outputs, and carryover-control variables documenting whether a Stage 5 stimulus matched the participant’s originally selected Stage 4 stimulus.本数据集支撑一项序列性实证研究，该研究聚焦用户在AI生成界面候选方案（AI-generated interface candidates）完成生成后，如何对其开展评估。本研究以AI生成的游戏主界面候选方案作为基于线索的评估对象，探究经过验证的感知线索如何与偏好选择、选择后审美评价、感知接受度及早期筛选判断建立关联。 本数据集包含三份匿名化的补充数据工作簿。 补充数据1（Supplementary Data 1）包含第三阶段（Stage 3）样本验证数据集与统计输出结果。该阶段共有52名参与者独立评估16个标记为S1–S16的AI生成游戏主界面试样，参与者需从五个感知线索维度对试样进行评分：视觉秩序（visual order）、视觉变化度（visual variation）、氛围丰富度（atmospheric richness）、生成界面可信度（generated-interface credibility），以及感知到的AI生成瑕疵（perceived AI-generated imperfection）。该工作簿包含试样级验证得分、信度结果、描述性统计，以及针对预设的视觉秩序×视觉变化线索结构的验证检查内容。 补充数据2（Supplementary Data 2）包含第四阶段（Stage 4）主用户评估数据集与衍生分析结果。该阶段共回收515份有效回复，参与者需从16个AI生成的界面试样中选出偏好的界面候选方案，随后仅针对选中的试样，基于简洁性、多样性、色彩丰富度与工艺感开展选择后审美维度评估。该工作簿包含匿名化的参与者级回复、标准化的S1–S16试样ID编码、试样类型映射、选中试样的线索映射、偏好选择分布、信度与结构充分性结果、线索-评分相关性、复合线索-评价关联、感知接受度分析，以及探索性的参与者级对比结果。S1–S16标签仅为中立的试样标识符，不对应MBTI类型、认知风格组别或人格分类。 补充数据3（Supplementary Data 3）包含第五阶段（Stage 5）后续机制检验数据集与混合效应模型（mixed-effects model）输出结果。该阶段从第四阶段的受访者池中选取200名自愿参与后续研究的参与者，每名参与者需评估8个代表性AI生成界面候选方案，最终得到1600条长格式的样本级观测数据。本数据集包含针对操作便捷性、视觉认知负荷、界面可信度、持续探索意愿与开发价值的条目级回复与量表得分，同时还包含试样元数据、从第三阶段映射得到的已验证线索得分、衍生线索指数、信度与描述性统计、相关性检验、混合效应模型输出，以及用于控制残留效应的变量——该变量用于记录第五阶段的试样是否与参与者在第四阶段原本选中的试样一致。