Authors:
加藤由樹、加藤尚吾

Title:
Anticipatory affective design in smartphone messaging: A within-AI test of human-reported affective-information availability in requests and invitations

Journal(書誌情報）:
Kato, Y., & Kato, S. (2026). Anticipatory affective design in smartphone messaging: A within-AI test of human-reported affective-information availability in requests and invitations. Computers in Human Behavior: Artificial Humans, 10, Article 100413.

doi: https://doi.org/10.1016/j.chbah.2026.100413

論文URL: https://doi.org/10.1016/j.chbah.2026.100413

Abstract:
Initial smartphone messages can be designed around what senders currently feel, what they intend to convey, and what they want recipients to feel before any reply occurs. We conceptualize this process as anticipatory affective design and examine it in requests and invitations. Human participants produced smartphone messages for interpersonal scenarios and reported their actual emotion, intended message emotion, and desired receiver emotion across nine emotion dimensions. These reports showed differentiated affective profiles, indicating that the three affective reference points were not interchangeable. We then tested, within a fixed AI generation architecture, whether message-level design differed when the scenario was supplemented with the same-case Human-reported affective-information block (A) rather than presented alone (N). Three stochastic generations per condition were planned for each of 234 retained cases. Message properties were assessed using source- and condition-blind coding supported by held-out measurement assessment with independent Human coders and an independent-provider LLM. Five prospectively eligible constructs showed positive A−N differences: Recipient Latitude/Response Optionality (Δ = .879), Imposition Acknowledgment/Burden Mitigation (Δ = .369), Positive Relational Affiliation (Δ = .231), Sender Affective Elaboration (Δ = .697), and Recipient Support Value/Interpersonal Reliance (Δ = .517); all Holm-adjusted p values were < .001. These findings show that, under the specified architecture, adding bundled case-linked Human-reported affective information altered observable message design across multiple interpersonal dimensions. The contrast identifies the total prompt/information effect of adding this block, not separate effects of its affective components, a Human–AI source difference, or recipient outcomes.

著者Contact先の email:
y-katou[at]star.sagami-wu.ac.jp（[at]を@に変更してください。）

投稿タグ