Who Gets To Be An Influencer?

A discourse on what it means to be an expert vs appearing to be an expert

Photo by Alex Iby on Unsplash

A discourse on what it means to be an expert vs appearing to be an expert

TLDR:

1. Social media platforms reward constant content generation
2. Speaking a lot isn’t the same thing as being worth listening to
3. The skill of attracting an audience isn’t related to the skills needed to build domain expertise

Introduction

The germination of this article came from something I noticed: many of the top writers on Medium in Artificial Intelligence don’t actually have a lot of experience working in the field. I did a deep dive into the AI influencers who write on Medium and found a case study that illustrates a larger problem with how social media works. It is too time-intensive for senior subject matter experts to build a large social media following.

Medium maintains a “top writer” designation for a number of subjects including Artificial Intelligence. There are at most 50 writers in any given category and they update it frequently, which drives a publish or perish dynamic.

35 of the top 50 writers in artificial intelligence were professional data workers, and my analysis will primarily focus on them.

Feeding the Beast

These 35 writers had a dizzying recent output of work. The following graph shows the number of articles each writer wrote in October of 2020, as well as the number of years of professional experience they have:

Image by author

Summary Statistics

Medium’s top writer algorithm is heavily weighted to output, which rewards prolific writers. This makes sense from Medium’s perspective, they want to promote writers who produce a lot of content on their platform to encourage others to do the same. Most social media platforms reward users who post regularly. Creating quality content at such a high rate of output is much easier to do earlier in your career.

Babies Teaching Babies

The first analysis I did was count the number of top AI Writers by the amount of professional experience they had.

There was a heavy bias towards writers with limited experience, a sharp decline when that experience passes the two-year mark, and a sudden resurgence in appearance around the 7-year mark.

Image by author

I think there are several things going on here:

1. Early career professionals are younger and have more time to produce the prodigious output required for attaining and maintaining the top writer status.
2. There is a lot of incentive for early career professionals to try and make a splash, especially if they’re attempting a career transition.
3. This looks a lot like a Dunning-Kruger Effect curve.

The Dunning-Kruger Effect

The Dunning-Kruger effect is a cognitive bias where people with lower amounts of experience tend to have much more confidence in their knowledge than people with substantially more experience.

Writing effectively about any subject requires the confidence to do so. Consequently, low experience people are much more confident in their knowledge because they don’t know what they don’t know.

If we bucket our AI writers by experience, grouping them into three categories, 0–2 years, 3–6 years, and 7+ years we can see a pretty good likeness of the Dunning-Kruger Curve:

Image from www.artandtechnology.com.au

Image by author

As data scientists get more experience, two things happen:

1. The incentives to write diminish, there is more focus on simply doing the job.
2. Confidence declines as data scientists increase in experience

Do we heed more experienced voices?

I also examined whether more experienced writers were able to attract more followers:

Image by author

While two experienced data scientists had the most followers by a healthy margin, there wasn’t a strong relationship for the rest of the set. Savvy junior data scientists could amass a substantial following.

I think this touches on the core problem with disseminating ideas in a social media setting. Social media companies like medium want to promote their most engaged users. In a field like artificial intelligence where experience really matters, especially when disseminating career advice; this creates a bizarre eco-system where professionally inexperienced influencers are dominating the conversation. This isn’t limited to medium AI writers but happens across social media platforms on a wide array of topics. For example, a top doctor on Instagram's followers are dwarfed by a fitness influencer.

Limitations

Obviously, we are looking at a small number of writers on a single platform. Twitter, LinkedIn, and YouTube are entirely different eco-systems.

Academic data scientists are far more concerned with publications and citations in peer-reviewed journals. This is true in a wide array of domains. Highly specialized experts are primarily interested in engaging with one another and less interested in engaging with the general public.

Data science as a new field will also skew towards a younger workforce, and people with deep experience are harder to come by.

Lastly, I didn’t evaluate the specific content the writers produced. Instead, I focused on the quantity of content produced and other easy-to-measure statistics. The data is certainly biased to what was relatively easy to measure and put together.

Conclusion

I have deep reservations about how much of the popular discourse around artificial intelligence is directed by inexperienced data scientists.

This leads to a number of perverse effects, especially when so much of the audience are aspiring data scientists themselves.

Artificial intelligence is a new field experiencing a gold rush, but most organizations aren’t in a position to be on the bleeding edge. Instead, they need more mundane work to build that future capacity. Unfortunately, this routine work doesn’t attract the same kind of attention from aspiring data scientists. This leads to a distorting effect; causing aspiring data scientists to over-index technical skills that won’t help them succeed in an actual data science role.

The loudest voices in the room are not necessarily the wisest. It’s easy to forget that on social media which is like a room where everyone is shouting to get attention.

Changes I’d like to see:

• For certain subjects, Medium should factor in experience to determining their top rankings. Focusing so heavily on output skews the conversation away from more experienced writers.
• Putting together a special publication that specifically seeks out the opinions of experts in various fields could elevate the discourse on the platform. Instead of relying on one person publishing constantly, spreading the work across numerous people would make the time commitment manageable.
• Only 4 of the 35 writers were women and I’d like to see more women as a part of this conversation. I’d like to see more gender parity across a wide variety of subjects.
• More generally I think social media platforms could do more to signal-boost actual experts. Some opinions are more valuable than others and would help fight misinformation by boosting real expertise

Are there any changes you’d like to see? If you like the status quo, what about it appeals to you? Do you think expertise should be a factor in how knowledge is disseminated on social media?

Notes

Of the 50 accounts, 4 of them were collective accounts and excluded from the analysis. I couldn’t find a LinkedIn profile for 3 of them, and 2 of them were full-time students (high school or undergraduate).

6 of the remaining writers were professional communicators, journalists, educators, and one gentleman who described himself as a futurist. I excluded them because they had a very different focus than the professional data workers.

I characterized the writers by the number of years of experience they had instead of by their job titles because some people inflate their job titles on LinkedIn, but it’s much harder to inflate years of actual experience.

All data for this article was recorded on November 8th, 2020. It’s likely that the data, such as the number of followers, and who is on the list have changed since then.

I decided not to publish the workbook to GitHub because it contains LinkedIn profile information for most of the writers and it doesn’t feel right to publish them in a public forum like that.

Demographic notes:

• 4 of the 35 writers were women.
• I didn’t track the ethnic or national demographics of the writers but it was a diverse set of writers from across the world
• 7 of the writers had PhDs
• I suspect 1 writer had substantially more experience than their LinkedIn profile revealed.

Self-deprecation:

I am entering my third year as a professional data worker and would categorize myself into the low-experience bucket.

I will never be able to hit the sheer output required to be considered a top writer because I just don’t have the time. My ideal publication schedule is one article a month.

About the author

Charles Mendelson is a marketing data analyst at PitchBook. If you’re looking for a guest for your podcast or YouTube channel the best way to get in touch with him is via LinkedIn

Originally published at https://charlesmendelson.com on April 1, 2021.