Why AI may overcomplicate answers: Humans and LLMs show 'addition bias,' often choosing extra steps over subtraction

2026-02-15 · Source: News on Artificial Intelligence and Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Social Sciences & Behavioral Studies · Depth: Fundamental Awareness, quick

Summary

Humans and Large Language Models (LLMs) exhibit "addition bias," a cognitive tendency to prefer solving problems by adding elements rather than subtracting them, even when subtraction would be simpler or more efficient. This bias, widely documented in human decision-making, leads to overcomplication. For instance, individuals might add more paragraphs to improve an essay instead of removing superfluous sections. The research indicates that LLMs also demonstrate this bias, suggesting a shared pattern in how both humans and advanced AI process information and approach problem-solving, often opting for increased complexity over simplification.

Key takeaway

For AI scientists developing or deploying LLMs, understanding addition bias is crucial. This shared cognitive pattern implies that LLMs might generate overly complex or verbose outputs when simpler solutions exist. You should implement explicit constraints or fine-tuning objectives that reward conciseness and penalize unnecessary additions, ensuring your models produce more efficient and direct responses, thereby improving user experience and computational efficiency.

Key insights

Both humans and LLMs exhibit "addition bias," preferring to add elements rather than subtract them for problem-solving.

Principles

• Cognitive biases influence decision-making.
• Addition bias favors complexity over simplicity.

In practice

• Review solutions for unnecessary additions.
• Prioritize simplification in problem-solving.

Topics

• Addition Bias
• Cognitive Biases
• Large Language Models
• Decision-Making

Best for: AI Scientist, AI Researcher, Research Scientist, AI Ethicist

Related on AIssential

• Language Models Exhibit Inconsistent Biases Towards Algorithmic Agents and Human Experts — LLMs exhibit inconsistent biases, preferring human experts in stated trust but algorithms in revealed preference tasks.
• LLMs displaying less cognitive bias are not necessarily better decision makers — Large language models (LLMs) are known to include both social stereotypes and cognitive biases, which researchers are actively working to identify, characterize, and rectify.
• I asked 5 different AIs to pick a number between 1 and 100… all of them said 42 😬 — LLMs exhibit cultural biases from training data, defaulting to non-random, culturally significant numbers when asked for randomness.
• Competing Biases underlie Overconfidence and Underconfidence in LLMs — LLMs exhibit both choice-supportive bias and hypersensitivity to contradictory information, leading to overconfidence and underconfidence.
• Less Is More: Cognitive Load and the Single-Prompt Ceiling in LLM Mathematical Reasoning — LLM mathematical reasoning hits a "single-prompt ceiling" where prompt complexity yields diminishing, unstable returns.
• 969: The Laws of Thought: The Math of Minds and Machines, with Prof. Tom Griffiths — Mathematical principles can describe both external physical laws and internal cognitive processes in biological and artificial intelligence.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by News on Artificial Intelligence and Machine Learning.