How to Make Your Multi-Image Posts Popular? An Approach to Enhanced Grid for Nine Images on Social Media

The nine-grid layout, consisting of a 3x3 grid of squares, is a common pattern widely used in various domains. In today’s social media, people are enthusiastic about showcasing images in a nine-grid format to express their appreciation for things and their love for life. Additionally, the nine-grid layout continues to find extensive applications in other fields such as user interfaces of mobile devices, game design, and information display. The arrangement of the nine-grid has a significant impact on user experience and usability. An effective arrangement can enhance user efficiency, visibility of information, and overall aesthetic appeal, thereby making the nine-grid more popular.

The nine-grid layout visually presents integrity, which is a special aspect of positioning images. Although there have been many studies on the impact of the order of images on user attention. However, there is still a lack of research on whether the order of images affects the user’s overall liking in the nine-grid situation. The following study will be based on two existing theories of the position significance of images to study the most popular image layout method among users in the nine-grid situation.

In examining the influence of different arrangement methods on the nine-grid layout, we have reviewed several relevant studies. Firstly, Keyan Ding et al[1]. aimed to develop a computational model for accurately predicting the viral potential of social images. Their study involved constructing a large-scale image database and employing a deep neural network-based computational model to analyze the individual contributions of image content to its popularity. However, their research primarily focused on scoring and predicting the popularity of single images, which differs from the focus of our study.Additionally, Hossein Taleb[2] proposed a method based on deep object recognition networks to reliably assess image quality with high correlation to human perception. This approach was employed for predicting image quality assessment and shares some similarities with our study, although the emphasis is different.

Furthermore, Ma Xiaoyue et al[3]. investigated the impact of image positioning and layout on user engagement behavior using multi-image tweet data from the Sina Weibo platform. They utilized an XGBoost model to predict the user engagement potential of images and conducted correlation and regression analyses, revealing that image positioning and layout moderately influence user interaction. This study explored the effects of different arrangement methods for multiple images, which overlaps to some extent with our research. Zhonghua He et al[4]. explored the relationship between the sequential position of images in nine-image grid and the Big Five personality traits. Their findings demonstrated that users can utilize prominent positions in the nine-image grid to highlight specific content, resulting in more captivating Weibo narratives. This study enriched the field of social media image research and revealed motivations for using multiple images.

In contrast, our paper specifically focuses on examining the impact of different arrangement methods on the nine-grid layout. Our novelty lies in exploring the influences of various arrangement methods on the visual appeal and narrative expression of the nine-grid images. By comparing different arrangement methods, we aim to provide specific guidance and insights on the optimal arrangement strategies for the nine-grid layout. Our research fills a gap in the existing knowledge and offers practical implications for designers and practitioners seeking to optimize the arrangement of images in the nine-grid layout.

There are two views on the position significance of images. The first one is the serial-position effect. It is the tendency of a person to recall the first and last items in a series best, and the middle items worst[5]. It was first documented by Ebbinghaus[6] in his seminal work on memory in the late 19th century, which is a fundamental principle of cognitive psychology. Numerous studies have affirmed its robustness across various tasks and stimuli types. While most research has focused on verbal materials, recent studies have extended this phenomenon to visual stimuli, including images.

The serial-position effect manifests as two distinct phenomena: the primacy effect and the recency effect. The primacy effect refers to the enhanced recall of items presented at the beginning of a sequence. This phenomenon is thought to occur because early items receive more rehearsal and encoding time, allowing them to be transferred into long-term memory more effectively. Meanwhile, the recency effect pertains to the superior recall of items presented at the end of a sequence. This effect is believed to result from items being held in short-term or working memory, which is still active during recall tasks[7]. In the context of image sequences (or nine-grid layout), the serial-position effect can be used to enhance users’ attention by strategically positioning impactful pictures at the beginning and the end. This balanced approach ensures that important information is not only introduced early but also reinforced at the end, maximizing the likelihood of retention and recall.

The second definition of significant position of nine-grid layout adheres to visual attention distribution[4]. Some researchers suggest that users will first pay attention to the center of the screen, which makes the position in the middle of the image sequence important. The image that is most compelling or fits the purpose most can be put in the center to gain more visual attention from users. However, further studies on which of the two definitions has the highest level of attractiveness in the situation of nine-grid layout are still needed.

We conducted a survey to investigate the perception of social media platform users towards the nine-grid layout. Out of the 36 valid responses received, all participants indicated that they had encountered nine-grid image posts on social media platforms, with 75% of them stating that they frequently browse such posts.

Subsequently, based on whether the participants had experience in creating nine-grid posts, the respondents were divided into two categories. The group with editing experience accounted for a high proportion of 77.8%, among which 96% of the respondents believed they possessed a certain level of arranging ability, while 93% stated that arranging would consume a significant amount of their time.We further investigated the preferences and experiences of this group in arranging the nine-grid layout,and the statistical results are shown in the Figure 2.The survey results revealed that for the most satisfying image, people tended to prefer placing it in the P5 position (96.43%) and the P1 position (32.14%), with positions in the four corners (P1, P3, P7, P9) being more popular than those on the edges. As for the least satisfying image, it was widely accepted to place it in the P7 position (39.2%), the P8 position (28.57%), and the P9 position (50%).For respondents who only had browsing experience, we attempted to understand their level of attention towards different positions in the nine-grid layout. According to statistics, the P5 position and the P3 position garnered significantly more attention than others, with probabilities of 100% and 62.5%.

Lastly, we constructed pairs of images consisting of a rearranged nine-grid thumbnail and its original arrangement. Participants were asked to choose the image with better presentation without knowing the labels of the unfamiliar images(Figure 2).This can help us to understand whether the arrangement of images is effective in improving the overall visual effect of the nine-grid images. The average number of votes for the ranked images was 23.4, significantly higher than the average votes received by the original images, which was 12.6.

To understand the factors that contribute to the popularity of nine-grid layout formats and explore the practices, preferences, and perceptions of individuals’ actions when utilizing the nine-grid layout, we conducted semi-structured interviews with five participants. By examining their current practices, image selection criteria, perceptions of prominence within the grid, and underlying motivations for employing this format, this study aims to lay the groundwork for devising effective strategies for enhancing the popularity and impact of multi-image content on social media platforms.

We invited five individuals with prior experience on posting multi-image combinations on social media like Sina Weibo, WeChat Moments, Instagram, etc. to participate in semi-structured interviews. Each interview session began with participants presenting an example of their nine-grid layout, followed by a series of inquiries regarding their motivations for posting photos on social media, their typical posting routines, criteria for photo selection and evaluation, and their perspectives on the importance of specific positions within the nine-grid format. Finally, there was a free discussion part where they were encouraged to talk about their relevant experience and tricks they have used previously relating to this topic. Convenience sampling was adopted for participant selection, acknowledging the pragmatic nature of this sampling method within the scope of the study. No incentives or rewards were offered for their participation. Interviews were conducted individually via phone calls to accommodate participants’ schedules and ensure the confidentiality of responses.

Purpose for posting multi-image posts on social media. The primary purpose identified for posting multi-image posts on social media is documentation and sharing, as articulated by all participants, who emphasized terms such as “documenting life moments” and “sharing daily experiences.” Additionally, one participant highlighted the significance of making interpersonal connections through visually appealing images, viewing it as a means to engage with diverse audiences. The other saw this action as a way of expressing thoughts and emotions. One participant specifically underscored platform differences, noted that “posting original pictures on Weibo maintains their image quality, thus the purpose of this platform is primarily for archiving. On the other hand, posting on WeChat Moments is more about sharing daily mood and communicating with friends.”

Moreover, participants expressed satisfaction in encountering multi-image posts created by others online. One person expressed their enjoyment, stating, “I enjoy watching imagery posts, especially when those pictures are arranged in an artful way. I followed a few bloggers who post photographs and movie screenshots every day as their posts entertain and inspire me aesthetically. I can really feel the content creator’s passion for making something beautiful if they pay extra attention on image arrangements.”

When questioned further about their decision-making process for image arrangement, participants provided insights into their strategies. Two participants typically follow a chronological order, while another mentioned deviating from this order to create a compelling narrative behind the photos. One participant differentiated between two types of posts, detailing, “for travel logs, I organize photos in a chronological framework. While for pure aesthetic photos with person, they are arranged based on clothing colors or body postures.”

After obtaining scores from both the Aesthetics-Quality Model and the Intrinsic Image Popularity Assessment Model, we proceeded to rearrange the images in the nine-grid layout. Each model’s scores were used to determine the order of arrangement, resulting in four distinct schemes: Aesthetics-Quality Model-based sequential arrangement, Aesthetics-Quality Model-based center prioritization arrangement, Intrinsic Image Popularity Assessment Model-based sequential arrangement, and Intrinsic Image Popularity Assessment Model-based center prioritization arrangement.

In the Aesthetics-Quality Model-based sequential arrangement, thumbnails were ordered in a reading order starting from the top-left corner, with the highest-scored thumbnail positioned first and the lowest-scored thumbnail placed in the bottom-right corner. This sequence aimed to maintain a logical flow for viewers’ visual perception.

Conversely, in the Aesthetics-Quality Model-based center prioritization arrangement, the highest-scored thumbnail took the central position, highlighting its significance. The following four thumbnails were placed in the four corners surrounding the central image, while the remaining four thumbnails were arranged in four lines around the central block. This layout intended to emphasize the highest-rated image while ensuring a balanced distribution of other images around it.

Similarly, the Intrinsic Image Popularity Assessment Model-based sequential arrangement and center prioritization arrangement followed the same principles but were based on the scores from the Intrinsic Image Popularity Assessment Model.

Each group resulted in two unique combinations, each formatted into a 900px × 900px image. These meticulously designed arrangements serve to optimize the visual impact and aesthetic appeal of image sets within the nine-grid framework. By providing data for subsequent user annotation, these arrangements lay the groundwork for further research into which model-based arrangement best aligns with user preferences.

Currently, there is not much research on nine-grid layout, so there are currently no publicly available nine-grid datasets. Therefore, we need to build the dataset ourselves, and the nine-grid layout in this dataset must be recently released. We crawled a total of 1,901 nine-gram charts published by 21 users on the sina weibo client as a dataset. These users come from a variety of fields, including film, design aesthetics, humanities, photography, stage design, and more. All of these users have more than 2,000 followers, including 17 users with more than 10,000 followers, 16 users with more than 100,000 followers, and 7 users with more than 1,000,000 followers. We chose users with more followers because users with more followers are likely to be more concerned about the quality of the nine-grid layouts they post, thus reducing the likelihood of getting low quality nine-grid layouts. We selected 250 of these nine-grid layouts that are representative of each domain . We arranged these nine-grid layouts in order of highest to lowest center layout according to aesthetic rating, highest to lowest order layout according to aesthetic rating, highest to lowest center layout according to content attractiveness and image quality, and highest to lowest order layout according to content attractiveness and image quality, so that the nine images in each original nine-grid layout form four layouts.

In order to test user preferences for the four nine-grid layouts, we used a publicly posted questionnaire to collect users’ preferences for the nine-grid images. The users who participated in the questionnaire were drawn from the entire community. We used a four-alternative approach, where users were asked to continuously choose their favorite arrangement from four different arrangements of the same nine-box grid.In order to reduce the interference of the position of the options, we designed different layout schemes for different questionnaires. Also to prevent user fatigue, we created 5 questionnaires with 50 questions in each questionnaire, and users need to answer all 50 questions before submitting the questionnaire. We received a total of 45 valid questionnaires, and the vote counts for the four arrangement schemes are presented in the table.

Overall, the arrangement scheme based on aesthetic quality outperformed the one based on visual content of the images, while the principle of center prioritization received higher evaluations compared to the sequential arrangement. Specifically, the combination of aesthetic ratings for individual images and the center prioritization sequence yielded the best results, whereas the scheme of sequentially arranging the images based on visual content ratings was noticeably inferior to the other schemes. Furthermore, both the sequential arrangement based on aesthetic ratings and the center prioritization arrangement based on visual content received favorable evaluations.

From the previous formative research, we have learned that although the first and center positions in a grid display are noticeably more likely to be prioritized over other positions, the center position also receives significantly higher attention than the first position. The statistical results from user annotations further support the evidence that a center prioritization arrangement scheme aligns with user preferences. This conclusion contributes to the understanding of human visual attention mechanisms and provides information for the development of theories and models related to visual attention and perception in psychology.

The process of human image perception is complex, as various factors such as color, brightness, clarity, subject, and background can influence people’s evaluations. Furthermore, different individuals may have different evaluations of the same image, which implies that the ratings provided by models may not be entirely accurate, leading to potential errors.In our experiment, we utilized the aesthetic quality scores provided by the NIMA model and the content ratings provided by the I2PA mode,both models demonstrated high accuracy and performance in their datasets. Therefore, we can argue that aesthetic appeal and quality of an image are more closely aligned with the way people evaluate images compared to all the visual information present in the image.