There’s lots of sector interest in storytelling and making it feel authentic.  Is a real story ever fake?

If it walks like a fake story and talks like one then the reader will feel it as such, whether it’s “authentic” or not.

On the flip side, AI generated stories, which are objectively fake, can feel real.

Stories involve our memory and there are two types to consider,

• Semantic Memory: Our memory for facts, concepts, and general knowledge about the world. It’s devoid of personal context.
• Episodic Memory: This is our memory for specific events and experiences, often associated with personal details and contextual information about “what”, “where”, and “when”.

There are many AI detection algorithms and the arms race to outsmart the detectors has only just begun.  What these algorithms look for to determine real or fake are instructive for making sure our real, human stories don’t feel like the latter.   This is only a partial list.

• Semantic Structures: AI-generated content primarily draws upon vast amounts of text and information, which is essentially semantic in nature.  Real stories include personal experiences – the “what”, “where”, and “when” details provide richness and context.
• Sentiment Analysis Disparities: Real stories have more varied sentiment, a range of emotions. AI tends to be more uniform in sentiment unless designed otherwise.
• Narrative Coherence: Real stories have meandering highs and lows.  AI can appear more straightforward or even formulaic unless designed otherwise.
• Detail Specificity: Real stories might mention very specific details vs. generic
• Repetitiveness: AI-generated text can sometimes be repetitive, especially over longer passages.
• Overuse of Common Phrases: AI models, due to their training on vast datasets, might overly rely on common phrases or idiomatic expressions.
• Emotional Tone and Nuance: AI-generated text might miss nuanced emotions or shift tones in ways that don’t feel natural to human readers. Detecting these subtle emotional inconsistencies can be useful.
• Semantic Density: Some AI models might pack sentences with more information or adjectives than typical human-written sentences.

But if I know how to sound real, authentic and human then I can prompt or fine-tune the AI model to be “real”.  And on the flip side, lots of real, human writing makes all the AI tell-tale mistakes and more, meaning your human writing lost its humanness.

An example, AI wrote both of these.

Version 1: AI-Produced

In Greenwood Park, there’s a tree that has been part of many memories. From sunny afternoons to moments of quiet reflection, this tree has seen years pass by. Today, it faces a challenge that requires our attention and care. With your help, we can ensure that the tree remains a cornerstone of the park for years to come.

Identifiers for AI:

• Generalized Descriptions: The AI version uses broader descriptors like “many memories,” “sunny afternoons,” and “moments of quiet reflection” without diving into specifics.
• Detachment from Personal Experiences: The narrative doesn’t involve personal experiences or evoke a sense of individual connection.
• Structured Flow: The progression from introducing the tree, to its importance, its current challenge, and the appeal for help is methodical and structured, which can sometimes be a hallmark of AI-produced content.

Version 2: Human-Like

My mom used to take me to Greenwood Park every Sunday, and we’d always sit under that one big tree, eating sandwiches and watching the clouds drift by. I was hoping to continue the tradition with my own kids. But now, that tree is in danger. It feels personal, like losing an old friend. If we come together, maybe we can save a piece of our shared history.

Identifiers for Human-Like:

• Specific Memories: The mention of a specific memory (sitting under the tree, eating sandwiches, watching clouds) anchors the narrative in a personal experience, making it more relatable and human.
• Emotional Connection: Phrases like “I was hoping,” “It feels personal,” and “like losing an old friend” add emotional depth and a sense of individual connection to the narrative.
• Conversational Tone: The phrasing and flow feels more like someone speaking, which can give it a more genuine, human touch.

Setting aside legitimate ethical concerns about suggesting something that isn’t, is…

The reader will either feel it like a real story or not and it likely has nothing to with whether a human or AI wrote it.

Kevin