Should You Use AI For Screenplay Feedback?

Lately, there is an increasing tendency for screenwriters and creative executives to use AI tools as a valid substitute for humans reading and obtaining insightful, actionable feedback on film and television scripts. Can a machine really capture the emotions and a screenwriter’s soul and intentions as well as a human being? Can you even tell the difference?

The nature of creativity relies on its humanism more than algorithms to be successful. How many times have you been given recommendations based on what you previously liked or what many people liked? How many of those recommendations didn’t appeal to you at that time, if at all? Before you answer, consider how often you agree with film and TV recommendations based on what you and your friends have previously watched? These suggestions are useful from a mathematical perspective, but can AI make suggestions based on the day you’ve had?

Tastes in stories are subjective. They are also evolving and constantly changing. Old trends may re-emerge and breakout hits may forge new trends. AI is incapable of harnessing these nuances. It only processes numbers.

AI relies on big data to draw conclusions, so it’s incapable of generating specifically-targeted feedback. Despite exponential improvements in AI capacity, the technology currently has a limited understanding of human emotions unless they are specifically spelt out. How many times have you asked yourself, “That machine knows me better than my best friends?”

It also can’t adequately process script components such as pauses, gestures, deflecting, or not responding to a comment to convey a script’s intent.

AI needs to generate a script rating based on comparative date in its database – i.e as many screenplays as it can process and the notes made on them. Can it handle multiple, sometimes opposing viewpoints, as are often made by creative executives? If so, who win? How might it handle a multi-verse screenplay like Everything, Everywhere All At Once or how does it decide how funny a comedy script is? How does it know if a joke needs punching up? How does it process cultural, gender, age, and other social considerations?

How does AI understand a writer’s background and perspective beyond what’s on the page? A war story written by an active serviceperson might have a different level of authenticity and impact than one written by an academic who’s never served.

Despite the lofty claims of AI eventually replacing the need for human readers, there is scant evidence for this. At best, it’s a tool that provides some generic framework or broad comments that a human reader may or write, but there is often little substantive critique in them. AI is useful for macro-assessments of scripts, but not micro-assessments. It cannot help a screenwriter win an Oscar.

Here is an example of some AI-generated overall feedback for an espionage thriller and how it compares to the human-generated version of an actual screenplay:

AI-Generated Feedback

This is a masterfully crafted script that seamlessly blends elements of espionage, personal struggle, and high-stakes action. The story follows a CIA operative as he navigates a treacherous world of secrets, betrayal, and the fight to save his wife from danger.

The script excels in its ability to create a sense of tension and suspense, keeping the audience on the edge of their seats throughout. The pacing is expertly handled, with each scene building upon the last, leading to a climactic and satisfying conclusion. The dialogue is sharp, and engaging, effectively conveying the complex emotions and motivations of the characters.

Human-Generated Feedback

The script is an adrenaline-infused action thriller that masterfully marries high-stakes espionage with a deeply personal tale of resilience and unyielding love. The script is a cinematic tour-de-force, taking us on a riveting journey through the life of an elite CIA operative whose life takes an unexpected turn when his beloved wife is abducted.

As he fights against the ticking clock and the undercurrent of political intrigue within his agency, he embarks on a quest to save her, uncovering along the way that she is not just a victim, but a woman of strength, grit, and uncompromising courage. While imbued with tension-filled action sequences that maintain a gripping narrative, the heart of the story remains firmly centered on its dynamic characters, making this a memorable exploration of love, sacrifice, and duty in the face of overwhelming odds.

Comparison

The AI-generated feedback feels too generic that it could be composed from a randomly-selected rolodex of comments cut and pasted from another screenplay. This is especially apparent in the second paragraph where the comments aren’t specific enough to the script and therefore difficult for the writer to know what to do next to improve their script.

The similarities are that both praise the quality of the script and point out its key features. Note how the human-generated AI feedback more accurately points out the emotional stakes of the CIA operative after his wife is abducted. It captures the urgency of the situation as the operative faces a time lock to rescue his wife which the AI-generated feedback doesn’t. Instead, it focuses on the mechanical aspects of the plot. Unless there are plot holes, or elements that require clarification, these comments are simple observations and difficult to action.

The human-generated feedback captures more character elements that can really elevate a screenplay. The operative is a man of principal, patriotism, and honor who serves his country to his best ability. It describes his struggle, his dilemma, and grit as he defies the prevailing odds to hopefully succeed in a corrupt political landscape.

There is also a perceptive evaluation of his wife’s character. She isn’t simply a damsel in distress waiting to be rescued. She has her own agency, resilience, and resourcefulness too. These comments can attract A-list talent because the character has multiple dimensions they can play.

Verdict

Currently, AI can be thought of as a broad diagnostic tool at best. The technology currently can’t offer specific, actionable suggestions to both improve a script. It isn’t capable of understanding deeper, abstract, philosophical concepts.

AI only compares word and action patterns in a script to what it’s been taught. A car chase might suggest an action movie, or a raised voice might indicate that someone is deeply emotional. But that’s about it. It’s much like learning a new language. You can learn words and phrases, but there is context, tonal inflections, and speech cadence that only native speakers can understand.

The final verdict is that the human touch is more useful in raising your screenplay from good to exceptional status. Use it!

For quality human feedback, go here: ISA Notes

or here: ISA Feedback