Part 5 of 5: This article is the fifth and final in a series prepared for The 13th Annual International Conference on the Image, “Here Comes the Metaverse: Designing the Virtual and the Real.” My presentation is entitled: Seeing is Believing: The Ethical and Legal Ramifications of Digital Image Manipulation.

---

“The more corrupt the state, the more numerous the laws.” — Tacitus, The Annals of Imperial Rome

Perhaps you agree with the tenet that the more laws a society passes, the more immoral that society. As an attorney, I find the principle true. If people didn’t litter, we wouldn’t need ordinances preventing it. If people didn’t steal, we wouldn’t need statutes preventing it. If people didn’t steal from their elderly parents, we wouldn’t need undue influence laws. We don’t need laws to protect the viewer if people don’t use image manipulation dishonestly.

In this final article of the series, (Part 1, Part 2, Part 3, Part 4) (alternative links to Part 1, Part 2, Part 3, Part 4) I propose, therefore, that as creators and consumers, we take personal responsibility for ethical image creation and manipulation detection. This is one person’s opinion. Mine. I call upon my peers and betters to develop and implement responsible solutions across industries.

DEFINITIONS AGAIN

If we rest on the principle that images are data, only simple and necessary adjustments are acceptable to retain an image’s photo designation. Anything that causes an image to become grossly dissimilar from reality — that violates the Too Many Nuts standard, results in an image designated as an illustration or digital art.

This approach aligns with many professional guidelines and passes expert tests, like these from the Stanford University Digital Ethics Project:

1. The Viewfinder Test: Viewers believe the image is what the photographer saw through the viewfinder.
2. Photojournalist’s Process Test: When a manipulation technique is beyond what a viewer understands can be done to an image, the manipulation is unacceptable.
3. The Technical Credibility Test: Level of technicality is not a benchmark. Even the most extreme acts of manipulation can go undetected.
4. Pregnant Bruce Willis Test: If the manipulation is not obvious (implausible), the burden is on the creator to disclose the manipulation.
5. Essence of the Image Test: Disclosure or obvious notice must accompany the image.

PROFESSIONAL ETHICS

Photojournalism

Peter Van Riper: If you can’t do it in the darkroom, don’t do it here.

In 2014, Doctor David Campbell, under the direction of World Press Photo, completed a research project entitled The Integrity of the Image. The goal was to ascertain — not to enforce — industry standards. He interviewed 45 experts (photographers, editors, and media executives) from 15 countries and concluded:

Manipulation, always unacceptable for news and documentary purposes, is globally understood to involve material changes to the image through the addition or subtraction of content designed to deceive or misleadthe viewer. Notice the highlighting of intent. Because most images are digital, and require processing from the raw version to become an image, the darkroom analogy is no longer a useful measure. Adjustments, like cropping, dodging and burning, toning, color adjustments, conversion to black and white, are acceptable as long as those adjustments are incidental and not excessive (or deceptive).

These conclusions inspired and underpin several professional codes of ethics and media publication standards. For example:

The Associated Press: AP pictures must always tell the truth. We do not alter or manipulate the content of a photograph in any way. The content of a photograph must not be altered in PhotoShop or by any other means. No element should be digitally added to or subtracted from any photograph. The faces or identities of individuals must not be obscured by PhotoShop or any other editing tool. Only retouching or the use of the cloning tool to eliminate dust and scratches are acceptable. Minor adjustments in PhotoShop are acceptable… (but) … Changes in density, contrast, color and saturation levels that substantially alter the original scene are not acceptable. Backgrounds should not be digitally blurred or eliminated by burning down or by aggressive toning.

National Press Photographers Association: Editing should maintain the integrity of the photographic images’ content and context. Do not manipulate images or add or alter sound in any way that can mislead viewers or misrepresent subjects.

Reuters: No additions or deletions to the subject matter of the original image. (thus changing the original content and journalistic integrity of an image). No excessive lightening, darkening or blurring of the image (thus misleading the viewer by disguising certain elements of an image). No excessive colour manipulation. (thus dramatically changing the original lighting conditions of an image). Only minor Photoshop work should be performed in the field (especially from laptops). We require only cropping, sizing and levels with resolution set to 300 dpi. Where possible, ask your regional or global picture desks to perform any required further Photo-shopping on their calibrated hi-resolution screens….

Society of Professional Journalists: Never deliberately distort facts or context, including visual information. Clearly label illustrations and re-enactments.

The National Union of Journalists offer the Not a Camera symbol to differentiate between a photograph and manipulated images:

In summary, these ethical guidelines put the burden on the photographer or creator to disclose the accuracy of the image, removing or reducing the possibility of deceiving the viewer.

ADVERTISING

Critics argue that the advertising industry does not self-regulate or disciplinethe socially harmful behavior of its members. Economic incentives negate the likelihood that advertisers will implement ethical guidelines. There is too much profit in narrative and fantastical imagery that motivates people to buy through emotion, sex, and aspirational appeals. And advertisers deem fashion and product photography as artistic illustrations — not documentary imaging.

However, some companies have taken a step to reduce or eliminate harmful advertising. To combat unrealistic body images and negative health effects, as of 2018, CVS eliminated all airbrushing to sell beauty products. Both Target and Dove have implemented body-inclusive advertising and product lines. By patronizing these companies, and boycotting companies that continue to manipulate images, consumers can support, and perhaps inspire, these ethical steps. However, without a clear we are not manipulating our images statement, consumers will be hard pressed to know who to patronize and who to boycott.

As false advertising and unfair competition laws, which regulate commercial speech and protect consumers, do not interfere with free speech, government intervention can be the primary means to control unscrupulous advertising images. (prior article)

TECHNOLOGY

Perhaps we can cure the ills of technology with technology? I’m skeptical.

Experts are developing networks to distinguish between real and fake video. Benjamin Penchas and Marco Monteiro have offered Sherlock AI, which detects deepfakes. Perhaps these forensic approaches can be expanded to include detecting manipulation of images?

Arguably the cause of powerful and realistic image manipulation with its Photoshop and Lightroom programs, Adobe is offering new technology to detect image manipulation. The technology examines noise statistics: metadata that reveals manipulation.

For now, we can rely on fake image detectors:

1. FotoForensics. Identifies compression levels. Significant variation indicates modification.
2. JPEGsnoop. Examines metadata and compression levels.
3. Ghiro. Detects edits through compression analysis.
4. Forensically. Detects manipulation through noise, error, component and clone analysis.
5. Amped Authenticate. Paid software which examines alteration in the image.
6. NeuralHash. Embeds unerasable digital watermark which directs back to the original image.

Fred Ritchin’s Four Corners Method suggests that all metadata would be accessible by placing the cursor over digital images. Users could immediately determine who owns the copyright and who made the image, what edits were made, what was cropped, and so on. Impractical for print media, we must rely on the evolution to an entirely digitized world. We also must reflect that metadata is not immutable. Perhaps in that not so distant future, clicking the image would allow the user to access the original, unedited image.

PERSONAL RESPONSIBILITY & MEDIA LITERACY

I found the following post in reaction to deepfakes:

I’m not sure what to say. The title almost says it all. I think the way these things are compared to “involuntary pornography” is completely absurd. The argument for banning them seems to rest on a premise that I fundamentally reject: that you own your likeness. But you don’t — I can take a picture of you without your consent and use it for whatever purpose I want. So, these things shouldn’t be banned, let alone outlawed — they’re just a natural progression of technology. In 20 years I’ll be able to have virtual reality sex with your wife, or your teenage daughter, or YOU, or whatever, and you won’t be able to do a damn thing about it, and that’s as it should be.

The writer’s deep misunderstanding of the law and legal rights is alarming. (So is the writer’s handle: butt_throwaway1.) The writer did clarify that we, as a society, must get to where we realize that video (images) are not real. Very pessimistic, but probably a wise approach for most of us. Towards that reality, many countries have recognized media literacy as a vital skill, both for an informed electorate and consumer protection, and have launched campaigns to educate their citizens.

As early as 2017, researchers found that most people cannot identify inconsistencies in a manipulated image unless it is painfully obvious. If you are curious whether you are good at identifying manipulated imagery, check out the posted testused for the experiment. I got a 70% accuracy. Perhaps I am more skilled than most at identifying image manipulation because I manipulate images for a living! The test, however, even for my pseudo-expert self, was difficult. You would think I would have gotten 100%. But I didn’t. And studies confirm that even experts have a difficult time identifying pure from manipulated images.

I have also had difficult discussions with students and teens who do not understand taking a screenshot of a photo or downloading an image is copyright infringement. The gross disregard of the purity and ownership of images is astounding. While we have made technological advances, we, as a species, have not made similar educational advances. Perhaps we should add an image module to civics courses (not that any school takes the time to teach civics)?

While we lobby for these educational steps, we can rely on the following methods to analyze images:

1. Reverse search. Perform a reverse search of the image to determine if alternative versions are exact or variant. You can also determine if a photojournalist or social media poster used an outdated photo to represent a current condition.
2. Check the metadata. You will need a metadata viewer (or this metadata viewer) or a Photoshop edit detector. The metadata will tell you who captured the image, when it was taken, using what camera and lens. However, this method is not foolproof — experts know to alter the metadata when they manipulate the image.
3. Look for reversed text. While we are mentioning reversal, neophyte manipulators often flip images when creating a composite.
4. Check the actual place. Like I did with Mister Faux Photographerrevealed in the first article in this series, if you know the place depicted, you can determine if the image is a faithful representation or a fake.
5. Examine for consistency in lighting. Are all the objects and subjects in the image similarly lit? If the light is coming from the right and the person in the middle has a shadow on the right side of her face, you should be suspect. You can also check for color cast: Each object or subject in the image should appear similar. If the guy at the end of the row appears to have green skin while everyone else looks natural, you may want to suspect he’s a lizard-person, but it’s just image manipulation.
6. Examine the shadows. Where is the light coming from in the image? Follow and determine if the shadow falls in a natural location. Or note if a shadow that should exist does not.
7. Examine reflections. Similar to shadow examination, this point is one of my go-tos for selfies. All those would-be supermodels in front of mirrors who cut away body girth — but forget the mirror is reflecting the truth.
8. Learn to use the compression technology listed above.
9. Get closer. Zoom into the image and notice any indication of merging, stacking, or pasting cloned items.
10. Examine the borders of the objects in the image. The edges of an element of the image should be clean and fit into the entirety of the image. If an element has been added or altered, the edges can appear jagged — or too smooth.
11. Check perspective. Photographers (and artists) see perspective errors quickly. For the rest of humanity, examine the image for objects that seem wrong. Tables are not at the same angle. Objects that seem to float. Objects seem too big or too small.
12. Check for other impossibilities. Like flawless skin. Or cat ears and whiskers on a human. Dead giveaway. Deepfakes, for example, slip as the person moves. Skin tone may change. And the subjects rarely blink naturally.
13. Look for warping. This indica will appear when the subject’s body part is increased or decreased. The body part might pass muster — but the background will warp.
14. Be wary of low-quality imagery. The use of low-quality images (blurry, pixilated) can indicate an intent to deceive. One author used an enlarged image of a video game to represent photos from the Ukraine-Russia conflict.
15. Look for remnants. Quick image manipulation often results in remnants of deleted objects. Ever see a ghost hand? There are three people in the image, but seven hands? Yeah. That’s a remnant.
16. Look for cloning. If you notice impossible repetition — the same bird in the same position in three places in the image, for example — you have some lazy image manipulation.

The Visual Social Media Lab provides a Visual Verification Guide to analyze images. The guide includes a helpful 20 Questions for Interrogating the Social Media Image. Through these methods, we can each carry the personal responsibility of separating photographs from fiction.

CONCLUSION

In conclusion, I would like to reflect on a straightforward and effective photo manipulation test proffered by NPPA Co-Chair and Past President, John Long. His three recommendations align with my own conclusions.

1. CONTEXT. Ask: In what context is the photo being used? Is the image for journalistic or documentary purposes? Little to no manipulation, but for production steps, is acceptable. Is the image art? Anything goes.
2. FAIR AND ACCURATE REPRESENTATION. Will the manipulation create an image that is grossly dissimilar to reality? Will the image pass the reasonable person Too Many Nuts test? (LINK) If not, return to the context question. If they produced the image for an art context, anything goes. For evidence or scientific contexts, the image must be unaltered. If the image is to be used for any other reason, answer question three honestly.
3. DECEPTION. Will the image deceive the viewer? What is the intention for making that image? Are we entertaining or reporting? In art, deception is expected. The viewer wants Houdini — to be entertained, enthralled. To be transported to otherworlds. Anything goes. But deception has no place in journalistic or documentary contexts. Nor should deception have a place in the political or social science arenas. And, perhaps we are learning that we must evaluate the fine line between advertising and art to avoid deception.

I conclude, no matter the context, the moment the image is no longer a photograph. If we hold these words dear — and use the term photograph to mean what they saw through the viewfinder is what they presented, viewers can once again trust photographs and photographers. Once the image is manipulated into a grossly dissimilar presentation, that image must be called digital art.

Yet, I offer this conclusion provisionally. I am one expert with one viewpoint resulting from my personal experiences. I invite my peers and my betters to engage, actively, in this dialogue and answer my question from my first article: What do we do with Mister Fake? 

---

References are embedded within the article, and you can peruse the complete list of resources here.