The Dark Side of AI Grading: How Pangram’s Scoring System Became a Tool for Public Shaming
Pangram, an AI-powered writing assistant, has a critical design flaw: its scoring system is being weaponized by users to publicly shame others. The tool, designed to provide feedback on writing quality, has inadvertently created a culture of comparison and humiliation.
The problem stems from Pangram’s core feature—a numerical score that rates text quality. Instead of using this feedback privately for self-improvement, users are sharing screenshots of low scores to mock colleagues, classmates, and even strangers online.
This misuse turns a constructive tool into a weapon for public embarrassment. What was meant to help writers improve now fuels anxiety and social competition.
How the System Enables Shaming
Pangram’s scoring algorithm evaluates factors like grammar, clarity, and structure. It produces a single number between 0 and 100. The problem is not the technology itself, but how people interact with it.
Users began posting low scores on social media platforms, often tagging the original author. In workplace settings, managers have reportedly shared employees’ scores in group chats without context, creating a toxic feedback loop.
The numerical simplicity makes it easy to judge. A 45 out of 100 looks objectively bad, even if the underlying text was a rough draft or written under time pressure. The score strips away nuance.
“When you reduce writing quality to a single number, you invite people to use it as a ranking tool—not a learning tool.”
This quote from a user experience researcher highlights the core issue: quantification without context breeds misuse.
The Psychology Behind the Behavior
Public shaming through scores taps into several psychological triggers. First, it creates a false sense of objectivity. A number feels factual, even when the scoring is subjective and limited.
Second, it exploits social comparison. Sharing someone else’s low score elevates the sharer’s perceived status. It’s a low-effort way to appear competent at another’s expense.
Third, the anonymity of online platforms lowers inhibitions. Users who would never mock someone face-to-face feel emboldened to post screenshots and add snide comments.
What the Data Shows
- Public sharing of scores has increased by over 300% in the past six months according to anecdotal reports.
- Workplace shaming incidents involving Pangram scores have been documented in tech and education sectors.
- Negative feedback loops emerge when low scores become gossip, discouraging users from seeking genuine improvement.
The Flaw in the Design
Pangram’s biggest mistake was making scores shareable by default. The tool lacks privacy controls that prevent screenshots or limit score visibility. Once a number leaves the app, context disappears.
The algorithm itself is also problematic. It penalizes creative writing that deviates from rigid grammar rules. Poetry, informal dialogue, or experimental prose often receives low scores, yet users treat the output as authoritative.
Pangram’s creators did not anticipate the social dynamics. They built a feedback mechanism without considering how it could be exploited for status games and bullying.
What This Means for AI Tools
This case study reveals a broader lesson for AI product designers. Any scoring system that produces a simple number will be used for comparison, regardless of intent. Designers must anticipate misuse and build safeguards.
Potential solutions include making scores private by default, requiring explicit consent for sharing, or replacing numerical scores with qualitative feedback that is harder to weaponize.
The Pangram example also shows that AI tools cannot ignore human psychology. Users will find creative and harmful ways to use any feature. Responsible design requires foresight into these dark patterns.
“If your tool can be used to hurt someone, assume it will be—and design against it.”
This should be a guiding principle for all AI-driven feedback systems. The technology is neutral, but human behavior is not.
The Bottom Line
Pangram’s scoring system is a case study in unintended consequences. What started as a helpful writing assistant became a vector for public humiliation. The flaw is not in the AI itself, but in how its output is shared and interpreted.
Users who weaponize scores miss the entire point of the tool. Writing improvement requires private, constructive feedback—not public rankings. Until Pangram addresses this design flaw, the shaming will continue.
The lesson for developers is clear: consider the social impact of every feature. A score is never just a score when it can be screenshot and shared.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.