Earlier Seen, More Often Chosen: Image Loading Speed as a Potential Seed of Hidden Manipulation

NordiCHI’26 | 2026.10.05 @Vaasa, Finland Earlier Seen, More Often Chosen: Image Loading Speed as a Potential Seed of Hidden Manipulation Itsuki Kanaya, Takaya Miyama, Yuichiro Kinoshita, Satoshi Nakamura （Meiji University） 1

Background [1/5] Focus on factors that can influence user choice These elements are visually noticeable In particular, we focus on: Choice-influencing factors in apparently fair interfaces 22

33 Background [2/5] Hidden Manipulation - Factors that influence user’s choices without being recognized as guidance. e.g.）Positions / Color / Font Style … [Valenzuela+ 2009] [Sekiguchi+ 2023] [Ueki+ 2023] Hidden Manipulation can be problematic ・Users may not realize their choices are being influenced. ・Designers may unintentionally affect user’s choices. We need to identify factors that can become Hidden Manipulation. Valenzuela, A., et al. : Position-based beliefs: The center-stage effect. Journal of Consumer Psychology, 19, 185–196, 2009. Yuto Sekiguchi, et al. : Does the Average Color Influence Selection?, HCII 2023, Vol.LNCS, volume 14012, pp.485-496, 2023. Riho Ueki, et al. : Does the Type of Font Face Induce the Selection?, HCII 2023, Vol.LNCS 14012, pp.497-510, 2023.

Background [3/5] Many choice interfaces use images as options. e.g., online shopping, mobile ordering, character voting etc. Hidden Manipulation could be problematic in these interfaces. 44

Background [4/5] Differences in loading speed of image can occur naturally. Due to factors such as network latency, File size etc. The loading speed can differ from image to image. → Some images may appear earlier or later than others. 55

66 Background [5/5] If loading-speed differences can bias user choice, they could be applied in situations like these: Online Voting – Characters, people etc. Online Purchases – Mobile ordering etc. Whether changing image loading speed could influence user choice

Research Question How does image loading speed influence user choice in an apparently fair interface? 77

88 Experiment Overview [1/3] Two experiments using two image presentation formats - Examined the effect of image loading speed on user choice Based on common image loading behavior on the Web Experiment 1：Appear from top to bottom Web Experiment Experiment 2：Gradually become clearer Web Experiment Loading speed did not affect the final appearance of each image.

Experiment Overview [2/3] Task：Select one image from four options - Questions about user’s preferences, as shown below 20 trials per participant - 15 main trials ， 5 attention checks 99

Experiment Overview [3/3] Speed Conditions Synchronous Earlier-Loading Later-Loading 10 10 -> Same loading speed for all four images -> Faster loading for one image -> Slower loading for one image 《Hypothesis》 Earlier loading increases selection in both formats. → Compare earlier-loading selection rate with chance（25%） Recruited 1,000 participants for each experiment via crowdsourcing.

Experiment 1 Images appear from Top to Bottom 11 11

12 12 Experiment 2 Images gradually become clearer ＜Blur＞ Blurred→Clear ＜Pixel＞ Pixelated→Clear

13 13 Experiment Result Selection rate of the target image with four options Exp. 1 Chance: 25% vs. Early-loading: 28.73% p < .01 （n=1,695） Exp. 2 Chance: 25% vs. Early-loading: 28.15% p < .01 （n=2,625） Blur: 28.34%，Pixel: 27.99% Earlier-loading options were more likely to be selected. → Hypothesis was supported

General Discussion [1/3] ・Earlier-loading option: 25% → 28% +3 percentage points A small bias can matter in large-scale or close decisions. e.g., elections, high-value purchases https://cigs.canon/article/20241028_8417.html The effect may be more relevant to casual choice situation. e.g., mobile ordering, online voting like MLB All-Star voting 14 14

• https://cigs.canon/article/20241028_8417.html

General Discussion [2/3] ・Potential for Hidden Manipulation Loading differences commonly occur naturally on the Web. →Similar differences can also be intentionally introduced. →Users may perceive them as natural loading behavior. Designers may unintentionally influence user’s choice. →Preload images to align when images become visible to users. 15 15

General Discussion [3/3] 16 16 Earlier-loading options were selected more often → One possible explanation：Primacy Effect Primacy Effect - Information presented earlier tends to have a stronger influence on judgement and choice. [Krosnick et al. 1987] Earlier-loading images became fully visible first. Earlier visual availability may provide a selection advantage. → The earlier-loading effect may be related to the primacy effect. J. A. Krosnick, et al.: “An Evaluation of a Cognitive Theory of Response-Order Effects in Survey Measurement”, The Public Opinion Quarterly, 51, 2, pp. 201–219 (1987).

Limitations and Future Work 17 17 ・Simple four-option selection task → More complex decision-making tasks ・Only limited loading-speed conditions were tested → Various realistic loading speeds ・Participants’ awareness of potential influence of loading speed on their choices was not measured. → Assess awareness of both loading-speed differences and their potential influence on choice.

Take-home message from our study Image loading speed can influence user choice. Avoid unnecessary loading-speed differences between images → Reduce unintended choice bias 18 18