RESEARCH FILM / ENGLISH

Distributed Creative Search on Digital Platforms

展开完整英文讲解

The same dish. The same creator. But a different camera angle, rhythm, or performance. When a creator tries something new, what actually changes?

Think of a video as three layers of building blocks. Conceptual content is what the work is about: the dish, the ingredients, and the theme.

Audiovisual production is how it is made: camera angles, color, sound, and editing. The same dish can be presented in a different way.

Embodied enactment is how the creator appears and performs: their face, movements, and posture. A familiar subject can be performed differently.

We studied ten food and cooking creators, using sixteen hundred and thirty-two videos with complete representations across these three dimensions.

How do we measure these layers? We turn video descriptions into text embeddings. Images, sound, and editing describe production. Face and pose sequences describe embodied performance. Normalized channels receive equal weights within each block, and the three blocks receive equal weights.

Change was usually uneven. In seventy-nine of ninety-five adjacent-period comparisons, one dimension carried at least half the squared movement, while another carried no more than one fifth.

Like renovating part of a house, creators changed some elements more than others. And the leading dimension could switch: the creator-balanced switch rate was thirty point three percent.

Which layer changed first? Among seven creators with classifiable first observed changes, six entered through audiovisual production. This is the beginning of our observed record, not necessarily their creative careers.

After a comparison without identified change, the pattern was different: four entries were audiovisual, four conceptual, and three ambiguous. There was no single fixed doorway.

For each local event, six earlier works form a baseline. We locate the next work, then ask whether its two successors continue in that direction. The peer comparison uses earlier works by other creators.

But trying something once is different from carrying it forward. Imagine a new editing style that appears in one video, then disappears from the next.

Now imagine that the next videos keep using that style, even as their dishes change. The adjustment has become part of what comes next. We call this continuation.

Across one hundred and six events, changes more concentrated in one dimension were positively associated with stronger continuation. The estimate remained uncertain: its nominal ninety-five percent interval included zero.

Stronger continuation was also positively associated with later differences from prior peer works, again with an interval spanning zero. This suggests a possible route toward distinctiveness, rather than establishing an inevitable process.

For the collective view, we average works within half year periods and compare the same creators across periods. Dispersion measures their spread around the group average. Each creator is also compared with their three nearest peers.

Now step back from one creator to the group. Each position represents a style of work. Nearby positions mean more similar works; greater distances mean larger differences.

If everyone takes the same step, everyone moves, but their distances need not change. Individual movement alone cannot tell us whether collective differences grow.

The observed creative space sometimes opened out and sometimes drew together, like a breathing balloon. Across eleven transitions, five expanded and six contracted. This is not a fixed cycle.

Dispersion change separates into two components: variation in creators' movements, and how those movements align with starting positions. Observed alignment was inward in all eleven transitions; expansion occurred when movement variation outweighed it.

Measurement uncertainty matters. Correcting for sampling error reduced the magnitude of average inward alignment by seventy-one point four percent. The five-six count remained, but four individual transition classifications changed.

The dimensions also played different descriptive roles. Production was the common first observed entry. In the frozen, uncorrected comparison, conceptual content showed the largest movement variation and opposing alignment; embodiment showed the smallest.

Changes could leave a trace. In eight out of nine comparisons spanning three periods, dispersion remained on the same side of its initial baseline at the third observation. It did not have to keep moving in the same direction.

Meanwhile, the nearest peers could change. These are neighbors in the similarity of their works, not necessarily friends or collaborators. Collective structure can persist while local similarities are rearranged.

Return to the building blocks. Ask which parts change, which changes carry into later work, and how those works compare with peers. Then look at the changing pattern across creators.

Our exploratory panel connects two views of creative search: how creators reconfigure their work, and how differences across creators evolve. Understanding creativity means following both what changes and what stays.

素材来源

以下作品片段用于研究说明；不作为已识别变化事件的证据。