AI-Behavioural Framework: We Are Only Humans

Have you ever wondered how AI can help us understand and influence human behaviour? How can AI learn from psychology and vice versa? How can AI help us achieve our personal and professional goals, cope with our challenges and emotions, and improve our well-being and relationships?

These are some of the questions that a new approach called AI-behavioural framework tries to answer. AI-behavioural framework is a term that can refer to different methods that combine AI and psychology to study or change human behaviour. For example, one possible AI-behavioural framework is the one proposed by Waken.ai, a research lab that offers a new AI-behavioural framework that merges psychology and engineering to shed light on whether AI, specifically Large Language Models (LLMs) like ChatGPT, could possess any level of consciousness.

According to Waken.ai their AI-behavioural framework consists of two phases: inception and introspection. In the inception phase, the AI system is given a surrounding context without any specific information and asked to imagine something related to it.

To try and explain this approach, I am going to use a metaphor of how a child losing his mother at 14 uses a process of inception and introspection to develop into adulthood, to try and guess what his mother would have wanted.

The child is like a chatbot that has been pre-trained on a large dataset of his mother’s words and actions. He has learned to predict what his mother would say or do in different situations, based on her values, beliefs, emotions, habits, and goals. He has also learned some facts about the world and some reasoning abilities from his mother’s dataset.

However, the child’s pre-training is incomplete and imperfect. He does not have access to his mother’s dataset any more, and he has to face new situations and challenges that his mother never encountered or taught him about. He also has to deal with his own grief, trauma, and identity crisis that his mother’s dataset cannot help him with.

Therefore, the child has to fine-tune his model on a narrower dataset that he carefully generates with his own introspection and imagination. He uses the inception phase to provide himself with surrounding context without any specific information. For example, he may ask himself “What would my mother want me to do in this situation?” or “How would my mother feel about this decision?” or “What would my mother say to me if she were here?” He then uses his imagination to generate possible answers based on his pre-trained model and his own intuition.

He then uses the introspection phase to analyse his internal state and compare it to the surrounding context provided in the inception phase. Likewise, he tries to determine if his answers are consistent with his mother’s values, beliefs, emotions, habits, and goals. He also tries to determine if his answers are helpful and appropriate for his own well-being and development. He uses feedback from himself and others to evaluate and improve his answers.

Through this process of inception and introspection, the child tries to cope with his loss and grow into adulthood, by guessing what his mother would have wanted for him. He also tries to discover and develop his own values, beliefs, emotions, habits, and goals, by learning from his mother’s dataset and his own experience.

Another possible AI-behavioural framework is the one that involves using AI to design interventions that can help people change their behaviours for the better, such as quitting smoking, exercising more, or eating healthier. This framework requires understanding the psychological factors that motivate or hinder behaviour change, such as attitudes, beliefs, emotions, habits, and social norms. Then, AI can be used to tailor personalised messages, feedback, rewards, or nudges that can influence those factors and encourage behaviour change.

AI-behavioural frameworks can have various applications and implications for different domains and stakeholders. They can also raise ethical and social challenges that need to be addressed carefully and transparently. For example, how can we ensure that the AI systems are accurate and reliable? How can we protect the privacy and security of the data and information used by the AI systems? How can we avoid potential harm or misuse of the AI systems by malicious actors or unintended consequences? How can we ensure the ethical and responsible use of the AI systems and their outcomes?

If you are interested in learning more about AI-behavioural frameworks and how they can help us understand and influence human behaviour, please visit Waken.ai’s website to learn more about their AI-behavioural framework and their book “I, AI” that explores the extent to which AI can achieve self-awareness and reflection through the lens of the Nemo’s Mirror Test.