Artificial intelligence (AI) has made significant advances, from powering autonomous vehicles to supporting medical diagnosis. However, one important question remains. Can AI pass cognitive tests designed for humans? AI has achieved impressive results in areas such as language processing and problem solving, but it has struggled to replicate the complexities of human thinking.
AI models like ChatGPT can generate text and solve problems efficiently, but when faced with cognitive tests such as Montreal Cognitive Assessment (MOCA), which is designed to measure human intelligence, they do not work as well.
This gap between AI technical outcomes and cognitive limitations highlights important challenges regarding its potential. AI is still not consistent with human thinking, especially in tasks that require abstract reasoning, emotional understanding, and contextual awareness.
Understand their role in cognitive testing and AI assessment
Cognitive tests such as MOCA are essential for measuring various aspects of human intelligence, such as memory, reasoning, problem solving, and spatial awareness. These tests are commonly used in clinical settings to diagnose conditions such as Alzheimer’s disease and dementia, and provide insight into how the brain functions in a variety of scenarios. Tasks such as recalling words, drawing clocks, and recognizing patterns assess the brain’s ability to navigate complex environments that are essential to everyday life.
However, when applied to AI, the results are completely different. AI models such as ChatGpt and Google’s Gemini may be superior in tasks such as pattern recognition and text generation, but they struggle with aspects of cognitiveness that require deeper understanding. For example, AI can complete tasks according to explicit instructions, but it does not have the ability to reason abstractly, interpret emotions, or apply contexts, which are core elements of human thought.
Therefore, cognitive testing serves a dual purpose in assessing AI. On the one hand, it highlights the strengths of AI in efficiently solving data processing and structured problems. On the other hand, they uncover important gaps in AI’s ability to replicate all human cognitive functions, particularly complex decision-making, emotional intelligence, and contextual awareness.
With the widespread use of AI, applications in areas such as healthcare and autonomous systems require more than just completing tasks. Cognitive testing provides a benchmark to assess whether AI can handle tasks that require abstract reasoning and emotional understanding, tasks that require central nature of human intelligence. In healthcare, for example, AI can analyze medical data to predict illness, but cannot rely on providing emotional support or understanding a patient’s unique situation to make subtle decisions. Similarly, autonomous systems such as self-driving cars often require human-like intuition, which is lacking in current AI models, to interpret unpredictable scenarios.
Using cognitive testing designed for humans, researchers can identify areas where AI needs improvement and develop more sophisticated systems. These assessments also help set realistic expectations about what AI can achieve and highlight where human involvement is still essential.
AI Limitations for Cognitive Tests
AI models have made impressive advances in data processing and pattern recognition. However, these models face major limitations in terms of tasks that require abstract reasoning, spatial awareness, and emotional understanding. Recent research that tested several AI systems using Montreal Cognitive Assessment (MOCA), a tool designed to measure human cognitive abilities, has revealed a clear gap between AI’s strengths in structured tasks and the struggle with more complex cognitive functions.
In this study, ChATGPT 4o scored 26 out of 30 and showed mild cognitive impairment, while Google’s Gemini scored 16 out of 30, reflecting severe cognitive impairment. One of the most important challenges of AI was visual spatial tasks, such as drawing clocks and replicating geometric shapes. These tasks, which require understanding spatial relationships and organizing visual information, are areas of human intuitive excellence. Despite being explicitly directed, the AI model struggled to complete these tasks accurately.
Human cognition integrates sensory input, memories and emotions, allowing adaptive decision-making. People rely on intuition, creativity, and context, especially when solving problems in ambiguous situations. This ability to think abstractly and use emotional intelligence in decision-making is an important feature of human cognition and thus enables individuals to navigate complex and dynamic scenarios.
In contrast, AI works by processing data via algorithms and statistical patterns. You can generate responses based on learning patterns, but you don’t really understand the context or meaning behind the data. This lack of understanding makes it difficult for AI to perform tasks that require abstract thought and emotional understanding. This is essential for tasks such as cognitive testing.
Interestingly, the cognitive limitations observed in AI models are similar to those seen in neurodegenerative diseases such as Alzheimer’s disease. In this study, when AI was asked about spatial awareness, its response was oversimplified, context-dependent, and similar to that of individuals with cognitive decline. These findings highlight that AI is excellent at processing and predicting structured data, but lacks the depth of understanding required for more nuanced decisions. This limitation relates in particular to health care and autonomous systems where judgment and reasoning are important.
Despite these limitations, there is a possibility of improvement. Newer versions of AI models such as the CHATGPT 4o show advances in inference and decision-making tasks. However, replicating human-like cognition requires improvements in AI design through potentially quantum computing or more advanced neural networks.
AI is fighting complex cognitive functions
Despite advances in AI technology, it has come a long way from passing cognitive tests designed for humans. AI is excellent at solving structured problems, but lacks when it comes to more subtle cognitive functions.
For example, AI models often miss marks when asked to draw geometric shapes or interpret spatial data. Humans naturally understand and organize visual information. This is a struggle for AI to do effectively. This highlights the fundamental problem. The ability of AI to process data is not equivalent to understanding how the human mind works.
The core of AI’s limitations is its algorithm-based nature. AI models work by identifying patterns within data, but lack the contextual awareness and emotional intelligence that humans use in decision-making. AI can efficiently generate outputs based on what they are trained, but like humans, they don’t understand the meaning behind those outputs. Coupled with a lack of empathy, the inability to engage in abstract thinking prevents AI from completing tasks that require deeper cognitive functions.
This gap between AI and human cognition is evident in healthcare. AI can assist in tasks such as analyzing medical scans and predicting illnesses. Still, human judgment cannot be replaced in complex decision-making that involves understanding the patient’s situation. Similarly, in systems such as autonomous vehicles, AI can process huge amounts of data to detect obstacles. Still, we cannot replicate the intuition that humans rely on in making momentary decisions in unexpected situations.
Despite these challenges, AI shows potential for improvement. New AI models are beginning to handle more sophisticated tasks, including inference and basic decision-making. However, even as these models advance, they remain far from the broad range of human cognitive abilities needed to pass cognitive tests designed for humans.
Conclusion
In conclusion, AI has made impressive advances in many areas, but there is still a long way to go before passing cognitive tests designed for humans. Although it can handle tasks such as data processing and problem solving, AI struggles with tasks that require abstract thinking, empathy, and contextual understanding.
Despite improvements, AI still struggles with tasks such as spatial awareness and decision-making. AI has shown promise for the future, particularly with technological advances, but it is far from replicating human cognition.
