Gender and Racial Bias in AI

A societal issue I read about was how Artificial Intelligence has a problem with gender and racial bias. The article also offers solutions in how to fix that. The article was written by Joy Buolamwini for Time magazine in 2019. Buolamwini is a computer scientist and poet of code who uses art and research to illuminate the social impacts of artificial intelligence. She has founded the Algorithmic Justice League, a foundation working towards creating more equitable and accountable technology in the world. The article itself is about her MIT Thesis, in which she used AI services from Microsoft, IBM, and Amazon, and used photos of famous black women to see how they AI would identify them. When she evaluated them, she found that for darker-skinned women, errors of guessing the gender of the faces were 35%, and failed to correctly classify the faces of Oprah Winfrey, Michelle Obama, and Serena Williams. She found that by using a white mask, the computer then was able to correctly identify the gender.


The issue surrounding gender and racial bias in AI, is due to the fact that women of color rarely are in positions to develop this technology. Most of the technology in AI fields are created by white males. The AI’s data set that it is using to help recognize faces also has less women of color than men or lighter skinned people. Buolamwini highlights the many different organizations that are trying to combat this bias in AI, and calls for governments and police to stop using this technology in identifying individuals, as it will incorrectly identify women of color and perpetuate a system of abuse.

How I’m Fighting Bias in Algorithms TED talk by Joy Buolamwini from March 2017

Looking Outwards 11: Societal Impacts of Digital Art

The article is about ImageNet Roulette, a classification tool that uses artificial intelligence to sort and categorize pictures. It was created by Trevor Pagan and Kate Crawford who were hoping to reveal some “racist, misogynistic and cruel results” with their platform to highlight issues with biases in artificial intelligence. The data set that trains the AI is widely used in the industry and consists of over 14 million images. The article highlights an extremely concerning issue of the AI highlighting white individuals with common descriptors regarding their careers, personalities, etc. However, black participants primarily reviewed results that reflected their race. Fortunately, changes were made to the database, and images and descriptors were removed to reduce the negative biases the software was creating.

https://www.smithsonianmag.com/smart-news/art-project-exposed-racial-biases-artificial-intelligence-system-180973207/

Looking Outward – 11: NFTs

Non-fungible Tokens, NFTs, gave the promise of increased income and a secure line of ownership to digital art creators. The idea was presented as a way for digital artists to reap the benefits of income and ownership that physical artists enjoy. This promised to elevate their work and to reduce copyright infringement by establishing a secure chain of ownership and proof of creation. Thus far, NFTs have failed to deliver on this promise, and in their current state, they are at best an experiment in the application of blockchain technology. At worst, they are a fad or even a fraud.

Jonathan Bailey outlines the rise and promise of NFTs in his article, “NFTs and Copyright”, which appeared in the March 16, 2021 edition of Plagiarism Times. Although NFTs have existed since 2017, they took the world by storm in 2021 with famous artists, such as Beeple, CEOs such as Twitter’s Jack Dorsey, and celebrities such as William Shatner, selling NFTs for millions of dollars. The word “non-fungible” means that something is not interchangeable, implying uniqueness. However, on the current NFT exchanges, anything can be tokenized, including URLs, images, and tweets. When something is tokenized a transaction is created on a blockchain. This makes the token unique, but only the token is unique. The original work that the token is linked to can be duplicated again and again.

Furthermore, the token does not confer ownership of the original work. When the token is sold, the original content can still be copied or owned by someone else. This creates ethical and copyright issues. These issues arise when people, or at times bots, tokenize and sell content they do not own. Some NFT exchanges attempt to prohibit unauthorized NFT sales, but many do not. This is copyright infringement and theft. A further ethical problem arises when NFTs are misrepresented to unknowing buyers. Many buyers believe they own the artwork or its copyright rather than just a transaction on a blockchain. Ultimately, it is a buyer’s responsibility to learn, however deliberate deception is ethically wrong.

NFTs could achieve their original promise and provide digital artists with income and protection; however, it would take a different implementation than what is now being pursued. In its current form, the NFT market will require artist and buyer protections or it will likely fade as another passing fad.

Project 11

Scrolling environment in space

sketch
//PLANET VARIABLES:
var numPlanets = 4;
var planet = {x: [], y: [], s: [], r: [], g: [], b: [], dx: []};
var d = [];

//STAR VARIABLES:
var numStars = 100;
var star = {x: [], y: [], s: [], dx: []};

//IMG VARIABLES:
var porthole;
var astroLinks = ["https://i.imgur.com/FrLKzou.png",
				  "https://i.imgur.com/vdhX4kE.png",
				  "https://i.imgur.com/01Kk3J7.png"];
var astroPics = [];
var astro = {x: [], y: [], s: [], dx: [], e: []};

function preload(){
	for(k = 0; k < 3; k++){
		astroPics[k] = loadImage(astroLinks[k]);
	}
	porthole = loadImage("https://i.imgur.com/YSSOdgW.png") //made this porthole graphic myself
}

function setup() {
    createCanvas(480, 480);
    background(0);
    planetInitialize();
    starInitialize();
    astroInitialize();
}

function draw() {
	background(0);
	starUpdate();
	planetUpdate();
	astroUpdate();
	image(porthole, 0, 0, width, height);
}

function drawPlanetA(x, y, s, i){ //option one for planet type (moon-y)
	strokeWeight(0);
	fill(planet.r[i], planet.g[i], planet.b[i]); //randomized color
	ellipse(x, y, s, s);
	fill(planet.r[i] + 20, planet.g[i] + 20, planet.b[i] + 20); //randomized color, but a little bit lighter
	ellipse(x - s/10, y + s/3, s/4);
	ellipse(x + s/5, y - s/10, s/3);
	ellipse(x - s/4, y - s/5, s/7);
}

function drawPlanetB(x, y, s, i){ //option two for planet type (saturn-y)
	fill(planet.r[i], planet.g[i], planet.b[i]);
	ellipse(x, y, s, s);
	strokeWeight(3);
	stroke(255-planet.r[i], 255-planet.g[i], 255-planet.b[i]);
	line(x - s*(2/3), y, x + s*(2/3), y);
	strokeWeight(0);
}

function starUpdate(){
	for(var j = 0; j < numStars; j++){
		strokeWeight(0)
		fill(250, 248, 235); //creamy white
		ellipse(star.x[j], star.y[j], star.s[j], star.s[j]);

		if(star.x[j] >= width + star.s[j]){ //if star has fully moved off screen, I reset the values
			star.s[j] = random(1, 10);
			star.x[j] = random(-20, 0-star.s[j]); //HOWEVER, I reset the values with the X position offscreen, so there appears to be a continuous scroll
			star.y[j] = random(0, height);
    	    star.dx[j] = star.s[j] / 200;
		}else{
			star.x[j] += star.dx[j]; //if star is not offscreen, it moves to the right
		}
	}
}

function planetUpdate(){
	for(var i = 0; i < numPlanets; i++){ 
		if(d[i] <= 1){ //selects planet type: if d is less than/equal to one, planet A is drawn, if d is greater than one, planet B is drawn
			drawPlanetA(planet.x[i], planet.y[i], planet.s[i], i);
		}else if(d[i] > 1){
			drawPlanetB(planet.x[i], planet.y[i], planet.s[i], i);
		}

		if(planet.x[i] >= width + planet.s[i] + (planet.s[i] * (2/3))){ //if planet has fully moved off screen, I reset the values
			planet.s[i] = random(10, 150);
			planet.x[i] = random(-200, 0-planet.s[i]); //HOWEVER, I reset the values with the X position offscreen, so there appears to be a continuous scroll
			planet.y[i] = random(0, height);
    	    planet.r[i] = random(20, 235);
    	    planet.g[i] = random(20, 235);
    	    planet.b[i] = random(20, 235);
    	    planet.dx[i] = planet.s[i] / 200;
		}else{
			planet.x[i] += planet.dx[i]; //if planet is not offscreen, it moves to the right
		}
	}
}

function astroUpdate(){
	for(var k = 0; k < 3; k++){
		image(astroPics[k], astro.x[k], astro.y[k], astro.s[k], astro.s[k]);
		if(astro.x[k] >= astro.e[k]){
			astro.x[k] = random(-2000, -150);
			astro.y[k] = random(0, height);
			astro.s[k] = random(30, 400);
			astro.dx[k] = astro.s[k] / 200;
			astro.e[k] = random(height+150, 2000);
		}else{
			astro.x[k] += astro.dx[k];
		}
	}
}

function planetInitialize(){
	for(var i = 0; i < numPlanets; i++){
    	planet.x[i] = random(0, width); //x position
    	planet.y[i] = random(0, height); //y position
    	planet.s[i] = random(10, 150); //size
    	planet.r[i] = random(20, 235); //r, g, and b are randomized. I seperated these instead of creating a color variable so I could use R, G, and B to edit the details
    	planet.g[i] = random(20, 235);
    	planet.b[i] = random(20, 235);
    	planet.dx[i] = planet.s[i] / 200; //dx is related to the size of the planet, if it's bigger it will appear to move quicker
    	d[i] = (random(0, 2)); //variable d selects whether or not planet type A or B is selected
    }
}

function starInitialize(){
	for(var j = 0; j < numStars; j++){
    	star.x[j] = random(0, width);
    	star.y[j] = random(0, height);
    	star.s[j] = random(1, 10);
    	star.dx[j] = star.s[j] / 200; //dx is related to the size of the star, if it's bigger it will appear to move quicker
    }
}

function astroInitialize(){
	for(var k = 0; k < 3; k++){
    	astro.x[k] = random(-2000, width)
    	astro.y[k] = random(0, height);
    	astro.s[k] = random(30, 150);
    	astro.dx[k] = astro.s[k] / 200; 
    	astro.e[k] = random(height+150, 2000); //astro end: beginning/end determines where image starts/ends it's journey before reset. I made the value larger so there would be greater diversity in when astronaunts appeared
    }
}

Looking Outwards 11

NFTs Are Shaking Up the Art World—But They Could Change So Much More

https://time.com/5947720/nft-art/ 

This article discusses the ways in which the art world could be altered via the growing popularity of NFTs (non-fungible tokens). It discusses many advantages, specifically positioning the creators of NFTs as down-on-their luck digital artists posed to receive notable benefits. These benefits include an increased ability to profit from one’s art, alongside the means to make one’s digital art pieces more exclusive, and hence, considered more valuable. This, however, completely neglects to mention the ways NFTs are actually used, alongside the culture surrounding them. NFTs are not bought because someone likes the art, or at the very least, that is not the primary reason for buying them. People buy NFTs to grow their own wealth at a gamble, and it’s safe to say that the already wealthy are not those who could lose the most from investing. Those who possess social capital (think CEOs, billionaires, celebrities) are easily able to construct their own NFTs and profit hugely, the reality being that very few digital artists are, in actuality, gaining financial stability from the practice, as the article implies. Like other assets on the blockchain, it would be ridiculous to pretend that NFTs don’t serve as a tool to further enrich the wealthy, usually at the expense of those outside of the 1%. Especially given the state of capitalism in this country, this is, without a doubt, extremely damaging. Though mentioned in the article, it is also worth re-acknowledging the fact that NFTs (and anything else on the blockchain) are bad for the environment, the computer clusters used to farm NFTs often powered by fossil fuels. This article is definitely biased in favor of NFTs, and perhaps unfairly fails to mention the realities of them. That being said, the title is right: NFTs could change so much more than the art world- change it for the worse.

Blog 11

NFTs have created vast spaces in which artists can get well deserved and long overdue compensation for their labor. Historically, digital art and media has been free for use on the Internet because of the high shareability and the ability to just copy and paste. NFTs encourage artists to be more creative as there is a space for reward and the Internet is limitless. However, NFTs do have a lot of societal and environmental downsides. Some works get plagiarized and sold, and copyright laws are blurred in the new space. NFTs also require a lot of compusing power and server farms are powered by fossil fuels. While there may be an assumption that online art is more environmentally friendly, it is actually quite the opposite.

https://time.com/5947720/nft-art/

Looking Outwards – 11

Impressions from women in media art

In Anna Gruaber’s “Women in Media Arts: Does AI think like a (white) man?” Gruaber features different artist/activists and their projects related to AI and feminism. Artificial intelligence is becoming a very important topic in regards to diversity and ethics. Activists have started pointing out problematic prejudices and distortions of supposedly objective algorithms. While there is a low proportion of women in IT professions, the real problem are the biased data sets used in AI and Machine learning.

Joy Buolamwini and Timnit Gebru are activists who investigate the prejudice of AI recognition systems. In this project, “Gender Shades,” the error rate is significantly higher among women; especially those with darker skin. Information about skin color is extremely important in the context of medical applications. 

Mary Flanagan’s project “help me know the truth,” shows that a discriminating algorithm does not come from the sexist or racist nature of the machine, but from the systemically racist structure of our society. “Help me know the truth” creates a perfect stereotype from a digital self portrait based on findings of cognitive neuroscience. 

Caroline Sinders wants to counterset bias. Her project, “Feminist Data Set,” is a multi-year art project that combines lectures and workshops to create interventions in the field of machine learning. Sinders wants to collect feminist data through them. The feminist data inclides artworks, essays, interviews, and books on feminism. The data attempts to introduce data collection as a feminist practice.

https://ars.electronica.art/aeblog/en/2020/04/10/women-in-media-arts-ai/

Looking Outward 11 – Racial Biases in Artificial Intelligence

In this article, Meilan Solly discusses a project (ImageNet Roulette) by Trevor Paglen and Kate Crawford that was created to expose the highly flawed and derogatory nature of AI human categorization. The project in question took the form of an AI driven identification tool that, when supplied with an image of a person, would return the category to which that image belongs (according to the algorithm). Categories or identifiers ranged on a spectrum of neutral to problematic terms like ‘pervert’, ‘alky (alcoholic)’, and ‘slut’.

While the category names are disturbing in and of themselves, the trends of categorization were far more so. Generally, the algorithm would sort people of color and women into extremely offensive categories at a disproportionately high rate. While no type of person was entirely safe from a harmful identifier, the disparity was clear. Solly describes a trial of the algorithm by a twitter user who uploaded photos of himself in various contexts and was only returned the tags “Black, Black African, Negroid, and Negro” (Solly). Paglen and Crawford have since removed ImageNet Roulette from the internet given that it has “made it’s point” (Solly), however it is still available as an art installation in Milan.

7-Training-Humans-24.jpg
ImageNet Roulette Installation

The implications of this project run deep. Arbitrary categorization of people on its own may have little consequence, but the underlying system to which it alludes is the same system that functions in the background of active AI processes with real world applications. Beyond this, the project makes comments on the concept of privacy online, having used thousands of images sourced from various locations without consent.

looking outwards-11

In Women in Media Arts: Does AI think like a (white) man?, Grubauer presents the view of Ars Electronica on AI for the purpose of supporting their project, which is to create “a comprehensive database devoted specifically to women in media art.” They did this to help give girls and women role models in terms of media art by increasing the presence of women artists in the public consciousness. Such projects play an important role in countering as hegemonic a phenomenon as patriarchy in the western world, which has shown through feminist philosophy and sociological research, that patriarchal (and white) tendencies permeate the cultural logic or societal common sense to the point of influencing objectivist science. “More and more activists point out the problematic prejudices and distortions of supposedly objective [my italics] algorithms. This has not only to do with the low proportion of female programmers and women in the IT sector in general, but above all with the biased data sets.” The mention of Mary Flanagan in this article aptly points out that this is a structural problem of society that creates the permeation I mention; it would be absurd to say an algorithm is inherently sexist or racist, and this claim is likely apprehended by confused defendants of the activists who push for making AI more equitable. The rest of the article introduces other women in the field of media art and their work such as Buolamwini/Gebru on skin color recognition, Aga/Manton on women-empowering AI, and Sinders’ feminist data set.

All quotes are from the article linked below.

https://ars.electronica.art/aeblog/en/2020/04/10/women-in-media-arts-ai/

Looking Outward – 11

The article “6 Art Projects Prying thre Lid Off Online Privacy” discusses the blurring of private versus public on the internet. We don’t realize how much of our data and information is shared without us even knowing. We use digital applications and services that are “free”, when in reality, data is the currency of social media. Who actually reads the fine print and understands what they are agreeing to? Many artists have explored this space in connection to identity, privacy, and data collection. One of the art pieces shared in this article was “fbFaces” by Joern Roeder and Jonathan Pirnay. Roeder and Pirnay used a crawler to search and find public profiles and collect images, facebook IDs, and names. They then pixelated these images, reflecting that they are no longer people, but a whole data network of information that people no longer have control over. This work fosters awareness of how data can be used and what this mean for privacy and identity.

Link : https://www.vice.com/en/article/4x4p43/6-art-projects-prying-the-lid-off-online-privacy