Jump to content


From Wikimania

Gretchen Andrew (born in California, 1988) is a Search Engine Artist and Culture Hacker who programs her paintings to manipulate & dominate search results.  She has completed projects or exhibitions with The New York Metropolitan Museum of Art, The V&A Museum, The Photographer’s Gallery, The British Film Institute, Cambridge University, Arebyte, The British Arts Council, The White Building, Ace Hotel, The London Film School, and Whitcher Projects. She trained in London with Billy Childish.


Art, Wikimedia, and the education of intelligent machines (and people too!)

Art is an underrepresented form of knowledge whose unique nature allows for information transmissions that are not possible in language. There are words that cannot be translated, histories, information and ideas that currently only, and maybe best exist, in the form of art. Ubuntu is one such word.  Its essence is in humanity, common recognition.

Art becomes even more important when we think about the next evolution of technology, thinking beyond just human readers and viewers, and considering the artificially intelligent machine viewers that use Wikipedia as a training ground. Through artificial intelligence, existing knowledge gaps are programmatically threaded into our future. So what would it mean for artificially intelligent machines to understand the word Ubuntu? Does the Wikimedia community understand its own role in the education of machines?

As long as existing policies such as such as Wikimedia Common’s COM:SCOPE actively discourage the inclusion of multiple art forms as a global kind of knowledge, we’re missing an opportunity to explore new forms of knowledge that can reach more of the world. Let’s do better.

This has been the focus of my own Wikimedia contributions, artwork and internet activism.  In this talk I will discuss new ways the Wikimedia community could think of art as educationally useful, including as a key educator of the machines already reading Wikimedia projects, and how changes in policy and outlook can be a doorway into new users and contributors, particularly with underrepresented communities.  Those attending will be encouraged to follow along on their own devices, seeing for themselves how restricting art-based contributions is contributing to systemic bias, and asked to consider how we might expand to include more kinds of visual knowledge in Wikimedia projects.

Background & Context: #SettleForBoth

I was recently at The Long Now Foundation where members were discussing, “What books would you want to rebuild civilization with?” Most of the contributions were scientific in their practicality with the great tombs of philosophy and government mixed it.  It’s a fun exercise that asks, “What knowledge is important?” But no number, size or weight limit had been placed on the library and I found that, without any constraint, the discussion was sort of useless.  Given the chance wouldn’t we want all books? All forms of knowledge? Even as an artist I’d not argue that Grapes of Wrath is more important than An Edible Guide to Plants but in this theoretical exercise, why not both (#settleForBoth)?

This is a form of the question facing Wikimedia Commons, a community whose purpose is tied to that of the Wikipedia Foundation's to encourage the, “growth, development and distribution of free, multilingual, educational content.”  Specifically Wikimedia Commons’ purpose is as a, “a media file repository making available public domain and freely-licensed educational media content (images, sound and video clips) to all. It acts as a common repository for all Wikimedia projects, but the content can be used by anyone, anywhere, for any purpose.”  It’s scope is defined with the restriction relevant to this discussion defined as, “Must be realistically useful for an educational purpose. The expression "educational" is to be understood according to its broad meaning of "providing knowledge; instructional or informative."

Purpose & Goals: Art's educational value as applied to the removal of "self made artwork without obvious educational value."

I’d like to make an argument for an expanded view of art’s educational value with the hope of having this view formally adopted by the community particularly as applied to the current practice of removing, "Artwork without obvious educational use, including non-educational artwork uploaded to showcase the artist's skills.”  

I make this argument in part because I have had my own images, most of them paintings, removed for this reason.  But the relevant context is much larger.  In making this argument I touch upon the purpose of art, art as knowledge, the relationship of art to language and art’s unique role in education while questioning the de facto supremacy of photography as truth’s and education’s medium of choice.  I also argue that this preference for photographic form perpetuates the internet’s product-focused content through which the internet’s great educational and creative potential is increasingly confined to a shopping mall.

Art & Knowledge Gaps

The question is whether or not my paintings, and others like them, constitute what Wikimedia Commons refers to as “obvious educational use.” The community's current interpretation of “obvious education use” is limited by a preference for text and photography that leads to a less diverse and therefore less complete set of contributors and contributions.  Especially as Wikipedia investigates the demographics of its contributors, one purpose of this writing is to make more obvious the educational use of art and its role in the Wikimedia Commons’s mission.

Artwork Marked For Removal as "non-educational."

On June 16th, 2017 53 of my images were marked for deletion off Wikimedia Commons as, “Artwork without obvious educational use, including non-educational artwork uploaded to showcase the artist's skills.”

The cultural/educational vs the commercial internet

If not to “showcase [my] skills” why was I putting these paintings on Wikimedia Commons? Compared to Google and Facebook, Wikimedia Commons does a decent job of reflecting the diverse and complex world in which we live.  For example, within Google “girl” means a sexualized woman but on Wikimedia the definition is more accurately reflected.  Search engines are getting much more content from sites that are trying to sell us something than sites that are trying to teach us something.

Companies Over Communities, Products Over People and Commerce Over Culture

Within search engines Amazon is more relevantly a company than a river. Cherokee more relevantly a car before a tribe of people. An internet made for discovery within Google will favor companies over communities, products over people and commerce over culture.  As more of the internet flows through Google and more content creators are incentivized to become relevant, companies with large search engine optimization resources outpace the ability of non-commercial concerns to surface. The results remind us to question: What makes up the internet, where does it comes from, who makes it and what are their motivations? Who is favored? Left out? Ignored?

Wikimedia Commons & Search Engine Optimization

Images on Wikimedia Commons are correctly structured for search engines in a user-friendly way and therefore have strong presence within search results.  This means Wikimedia Commons is playing an important role in the non-commercial internet through its ability to make content competitive within search engines.  This is especially important now as Wikimedia Commons is involved not just in the education of people but of machines.

Language & Images

If you feed enough photos of a platypus into a machine learning system it can learn to identify a platypus. It is one thing when you are discussing a word for which a definition is binary, something that either is or is not a platypus.  But what about man? Woman? House vs Home?  When conducting an image search when do you stop getting "yellow" and start seeing "orange", or stop defining as "black" and start saying "white", or "rich" and "poor", "beautiful” and “ugly?”

#accordingToTheInternet: human

When you search for “human” within images.google and see mostly white men, you can’t say that Google is technically wrong, these images are of humans.  This visual mapping of every word to a set of most relevant images is something old encyclopedias never would have attempted in part because the complexity of many words is failed by visual definition.

Through machine learning and artificial intelligence search engines like Google are learning to identify who is and is not a human based on the current, widely non-diverse data set of online content. Any holes and bias that exist within language are getting automated through machine education into how the world is being defined.  Putting content on Wikipedia Commons is potentially one of the most powerful ways to get diverse, complex and minority perspective into these definitions.  

Photography & "Truth"

We need to encourage and allow much more diverse set of images on high-value content sites such as Wikimedia Commons to aid in the education of machines, but why not just more photography? What is it about the presence of art-images, especially in non-photographic mediums, that makes them uniquely valuable in the education both of people and machines?

Search Engine Art: Internet Imperialism

My research has so far led me to conclude that one important approach is to add non-commercial, non-product images online that are structured in a way pleasing to search engines. Part of this initiative involves my method of Internet Imperialism where I create paintings about targeted keywords and then spread these images online with the express goal of modifying universally returned search results.  For example,  when you search “malignant epithelial ovarian cancer” Google now returns my paintings as top results.  

Painting & Perspective

It is not that within themselves my paintings contain the sole and correct definitions but that the painting medium, as compared to photography, is a great reminder of bias and perspective.  The potential for a diverse and complex interpretation is high. When we, and maybe so with machines, see a painting, and especially a painting within a search engine, we remember more easily that this images is one person’s perspective, not some techno-endorsed truth.  

By preferencing photography over other mediums such as painting and drawing we are drastically over estimating its relationship to truth. Photography too is a form of art through which conscious and unconscious bias are presented. Manipulating photographs is common and easy.  Computer generated graphics and photo realistic painting blur the confining lines of the medium.  Even medical diagrams often convey deeply seeded beliefs in their attempts at impartial truth.  

Art & Education

This is to say that my purpose for putting these removed paintings on Wikimedia Commons was not to, “showcase [my] skills.”  My purpose was educational.  Having these paintings on Wikimedia Commons allowed me to aid in the education of search engine algorithms, contributing some of the human complexity they are missing in their current product-focused training set.  They also contribute to the education of Wikimedia Commons’ and search engine’s end users.

Art constitutes a unique form of knowledge that often conveys its information through the productive introduction of uncertainty instead of the firming of facts.  When we look at a painting one of the important and often unconscious transmissions that occurs is that the way the world is seen and understood is subject to personal interpretation.  Through art we receive a jarring and expanding awareness of otherness.  Through art we can make machines less sure that they have a complete handle on what it means, and looks like, to be human.  Art can do the same thing for us humans.

Art is educational, outside of it’s contextualization by experts, in large part because through great art we gain a way of seeing the world that is not our own, that which by definition we could not have understood through our own direct experience. Art is the gift of experiences that we ourselves could not have had directly. Through “Starry Night” we get to know the world as Van Gogh knew it, and in that experience we are extended beyond the constraints of our own way of looking and by extension, understanding.

Subject vs Object & "better than words"

In the same way, Edvard Munch created paintings inspired by personal anxiety that we ourselves did not have to suffer, and in doing so he allowed us to feel, if only obliquely, what he experienced.  For examples like this, some in the Wikimedia Common’s community have argued for keeping images on a “better than words” basis.  

In Art this is the difference between a work’s objects and its subjects. In Munch’s The Scream we see a distorted human figure on a bridge, the objects, but read the subject to be anxiety or fear.  In another example, a painting of an apple (the object) may be about temptation (the subject).  Much of art’s educational value is in what it is about, in its subjects.  My images that were taken off of Wikimedia Commons were paintings about my hometown of Bow New Hampshire, what it was like to grow up there and return as an adult.  These images replaced and live among the top results on Google that were once dominated by for-sale real estate.  

The Difference Between "Of" and "About"

By allowing only images “of” instead of images “about” words on Wikimedia Commons we limit everything to its lowest common denominator and apply a lens that is fraught with bias on what the essence of something truly is.  When community deals with abstractions, such as emotions, artwork is highly present is submissions acceptable by the community.  This is likely because “sadness” has no image equalicant of “home.” But can not home be both a subject and an object? Isn’t home as a concept complexly related to cultural expectations and feelings? I believe this is an important place for art.  It says, “to the artist, this is a home” instead of, “this is what home is.”  This is central to art’s educational value.  

What is knowledge and how is it conveyed and by whom?

Ultimately to accept artworks more broadly on Wikimedia Commons is to more deeply consider what is knowledge and how is it conveyed. I believe that art, even poorly skilled attempts at art, tell us something essential about the uncertainty and complexity of things, the variety of ways the world can be seen and understood.

I think there is great complexity and diversity to be gained by the inclusion and encouragement of art within Wikimedia Commons. Especially as we consider its unique presence in the training sets being used by search engines in machine learning. & like with the books to rebuild civilization, why not #settleForBoth and consider these contributions, "useless by harmless."