ZINE
ZINE
Using AI To Quantify & Size META Trends
0:00
Current time: 0:00 / Total time: -10:52
-10:52

Using AI To Quantify & Size META Trends

Part I: AI vs. Human Rankings
Alexander Ant

Earlier this year Sarah DaVanzo and I published the fifth annual 2022 META Trend analysis, a distillation of 40+ industry trend reports to ultimately identify the most frequently reported (i.e. noteworthy) trends for the year. Fourteen (meta) trends were identified to represent what the entire trends industry was collectively forecasting.

During this time we also announced that for the first year, we’d quantify each of these META Trends to more precisely size and evaluate their cultural influence. It’s been a moment, but we finally crunched the data...

To get here though, we first had to answer a series of very thorny questions: How do we define the criteria or “borders” of each META Trend, which sources of data should be leveraged to score each, should we study these trends’ influence within a U.S. or global data set, how do we even collect and analyze this data at scale, and how do we complete this exercise without adding any human interference, tipping the scales of objectivity?

Once answering those could we then shine light on the larger questions at hand:

  1. By leveraging cultural data, would AI rank the META Trends differently than how the humans did?

  2. Left on its own, could AI identify similar or different (perhaps missing, overlooked) META Trends?

  3. And conclusively, what are the best tasks and roles for humans versus AI in order to develop a successfully orchestrated cultural intelligence system?

While asking ourselves all of these questions, we formed a partnership with the team at NWO.ai. The NWO.ai platform, which was amongst the finalists for LVMH's Innovation award, and an Industry Cloud partner of SAP, was formed in 2020 to identify consumer signals before they become exponential. Its AI algorithms learned over the last 2.5 years, and now boasts statistical trend prediction accuracy. They effectively quantify culture. With Sarah’s previous experience collaborating with them earlier this year, we determined their platform would be the perfect tool for our quantitative META Trend analysis.

Experiment 01.
2022 META Trend Scoring & Ranking
_ Humans vs. AI

As for our first question: If AI was to collect cultural data against each of the META Trends and then score their importance to rank each of them, would the AI ranking match or produce different results from the original human rank?

Answer: Different.

As a reminder, our “human ranking” was completed by Sarah and myself manually counting the frequency of similar trend mentions throughout the 40+ industry reports. For example, sustainability trends received the most attention and real estate across the analyzed 2022 reports, and hence the “Eco- Everything” META Trend was born and ranked in the top spot.

So, to more precisely size and rank these META Trends, NWO.ai’s AI calculated a series of keyword “portfolios” of each trend. These portfolios were essentially groupings of keywords and phrases (i.e. booleans) representing each of the 14 META Trends. Sarah and I authored these portfolios ourselves, but to curb any subjectivity, we only leveraged the language used from the original reports’ descriptions. To be clear, Sarah and I did not forecast these 14 META Trends – these were simply the most talked about concepts throughout the industry.

NWO.ai then measured the cultural importance of each META Trend via its portfolio of keywords. Consumer interest was quantified using a variety of data sources spanning: social, news publications, search, investments, patents, scientific journals, e-commerce data, and even film scripts. Ultimately, an AI-derived “Impact Score” was calculated by aggregating: volume (quantity of these signals across sources), frequency (volume over day), reach (distribution of publications), etc. These scores were finally normalized on a 0-100 scale to fairly pit each of the META Trends against one another and create our official AI ranking.

AI META Trend Scoring (Global vs. U.S.)

This AI ranking from the cultural data behind each of this year’s 14 META Trends revealed that the convenience economy (Now! Now! NOW!) is dominating culture by a magnitude of more than double that of some other META Trends. In other words, Now! Now! NOW! or the endless demands of innovation surrounding online shopping, has a cultural impact more than 3x the size of the META Trend xX~VIBES~Xx, which is our desire to tune in, drop out, and create spaces or purchase products to fulfill and focus.

With these AI scores, we then (re-)ranked each META Trend from the human approach.

META Trend Rankings (Global vs. U.S. vs. Human)

This is where things got interesting.

We had rank discrepancies.

While the human ranking process (i.e. mention count) identified Eco- Everything as the most prevalent META Trend... according to Eco- Everything’s portfolio of AI scored keywords, it is in fact the 9th most culturally impactful META Trend globally. Meanwhile Now! Now! NOW! received the highest Impact Score from the AI. Originally, it was only identified as the 10th most important META Trend. But according to the AI it is in fact #1.

Perhaps unfortunately we’re engaging in some wishful thinking here:

According to millions of cultural data points scraped and analyzed, consumerism beats out industry trend reports’ hype of sustainable innovation.

Experiment 02.
2022 META Trend Scoring & Ranking
_ U.S. vs. Global

When the first differentiation of rankings came back, we were only using a U.S. data set, achieved by geo-fencing our analyzed cultural data to North America.

We wondered if there would still be discrepancies in the ranking if we created a new, specific global rank by opening up our aperture, sources and data.

Was there a difference?

Answer: Not really.

AI META Trend Rankings (Global vs. U.S.)

The META Trend rankings by the AI are by and large the same globally vs. U.S., statistically reinforcing America’s cultural influence. In other words, META Trends’ impact in the U.S. reflects similar impact globally.

Therefore, this scoring suggests that U.S. trends can be proxies for global trends as they ultimately ripple outwards from the states.

But more importantly, because we tested the scoring twice (for the U.S. and globally), it reveals that AI data-driven trend scoring and ranking is impressively consistent.

So to zoom back out, the rank difference between the AI and human methods shows glaring variation. This should make us all think — perhaps even question — the subjectivity and accuracy of the industry’s reported trends. After all, we were scoring what the humans (the most experienced trend forecasters, no less) originally published.

How valid are these concepts if millions of data points and AI couldn’t mirror our collectively proclaimed importance of them?

Or conversely, maybe these were in fact the most worthy trends to score and we’re just splitting hairs between the most important of the important.

Or again on the other hand, just perhaps, there are trends out there with higher Impact Scores just never identified by the experts...

In any case... The AI scoring revealed that Now! Now! NOW!, Home Hubs and Radical Inclusivity are the top three META Trends of 2022, from both a U.S. and global perspective. This suggests that no matter one’s vertical, these three META Trends are both qualitatively (human-identified) and quantitatively (AI-scored) important.

Whether you’re national or global, strategically double down in these spaces.

Experiment 03.
AI Deconstructs The Anatomy of Trends
_ Identifying Drivers

In exploring NWO.ai’s platform, we noticed the AI could do something the human process could never — it allowed us to dig deeper and uncover the keywords (i.e. signals, concepts, trends, etc.) beneath the META Trend’s surface. In other words, what is driving each META Trend forward? If you recall, because we originally created portfolios of keywords for each META Trend to score them, we had the opportunity to score each META Trend’s DNA strands to determine which specific elements are having the most influence. Knowing what is driving a trend by ranking its most important components (i.e. keywords) can help us envision how it will evolve over time.

NWO.ai granted us the ability to plot each META Trend’s portfolio of keywords on a 2x2 matrix and rank them by their current growth, speed, tone and forecast. Essentially answering the question: which explicit keywords are growing or declining in volume, and is this change exponential or momentarily still?

Top AI-Identified Keyword Drivers of the META Trends

Conclusion:
AI Trend Scoring, Ranking & Driver Identification Is Superior to Humans

By leveraging AI to score each of the META Trends, not only did we create a more accurate prioritization with a U.S. vs. global nuance — something which humans could never achieve — but the AI also allowed us to reach a granularity and inspect the anatomy of each META Trend. We learned which components are most significantly influencing its growth.

But from these experiments, we also have a warning:

An AI data-driven system can prioritize completely different trends than us humans.

While AI can confidently unlock insight humans can only dream of, its results are so divergent from a human approach that healthy questioning is required: for the emerging software, but also primarily, the human’s “expert” input.

The AI declared that what we humans thought was most important was not actually the case.

This begged our next question:

What does the AI find most important?


Complete Series:
Part I: Using AI To Quantify & Size META Trends
Part II: How To Spot Trends with AI
Part III: A_Framework_To: Find Overlooked & De-bias Trends

3 Comments

Commenting has been turned off for this post
ZINE
ZINE
Alternative explanations and provocations around our tech, media, and overlooked social shifts. Webby-winning cultural intel for the curious.