Procrastination and AI: We'll Read This Later

In 1991, Nobel laureate George Akerlof explored a phenomenon many of us know too well: procrastination. His work, Procrastination and Obedience, introduced the concept of present bias, highlighting how we often prioritise immediate gratification over long-term benefits, even when it’s costly. Recent work by Chebolu and Dayan (2023) uses reinforcement learning (RL) to model procrastination as a decision-making process. Their findings suggest that people’s tendency to procrastinate is driven by a reward system that favors immediate relief over future gains—essentially, the same present bias Akerlof identified.? Fuschia M. Sirois (2023) proposed the Stress Context Vulnerability Model, which suggests that procrastination often occurs as a coping mechanism in stressful environments.

Today, as Gen AI backed tools (AI tools) like ChatGPT, Google Gemini and different types of Copilots or Agents become (slowly) part of our workflow, it’s worth asking: Can gen AI help us overcome procrastination, or could it actually reinforce Akerlof’s concerns?

AI tools can offer the potential to reduce task friction by helping automate and organise routine tasks. And if they are integrated in ways that alleviate workload stress—by assisting with information retrieval, scheduling, and organisation—they could, in theory, reduce procrastination. For instance, ChatGPT can draft content helping people initiate tasks they might otherwise delay.?

Yet, AI tools may unintentionally enable procrastination. People might rely on these tools for “quick fixes,” delaying more complex problem-solving and engagement with the work itself. That means “defer” real engagement with complex tasks, reinforcing patterns of procrastination rather than resolving them.

Why should we still leave out complex tasks? Well, as Mirzadeh et al. (2024) points out, Large Language Models (LLM) rely on pattern matching, not true reasoning. While they can assist with straightforward tasks, they lack the deeper logical reasoning that’s required for complex decision-making. This means that while AI tools can aid in “starting” tasks, they may not support the critical thinking needed to fully complete them and potentially reinforcing procrastination -at a later stage. rather than resolving it.

For the moment, a nice balance seems to be that we can use these tools to reduce routine workload and handle simpler tasks, freeing us to focus on work that demands our full engagement. Let's be mindful that the tendency to delay complex work won’t be solved by technology alone; it requires awareness of our decision-making biases and intentional structuring of tasks to promote sustained focus.

References:

1. Akerlof, George A. (1991). Procrastination and Obedience. The American Economic Review.
2. Chebolu, Sahiti & Dayan, Peter (2023). Reinforcement Learning and Behavioral Insights on Procrastination. Behavioral Science and Policy.
3. Sirois, Fuschia M. (2023). The Stress Context Vulnerability Model of Procrastination. Journal of Experimental Psychology.
4. Hofman, Jake et al. (2024). AI and Productivity: A Comprehensive Study on Copilot Usage in Outlook and GitHub. Microsoft Research Report.
5. Mirzadeh, Iman et al. (2024). GSM-Symbolic: Understanding the Limitations of Mathematical Reasoning in Large Language Models. arXiv Preprint.