The Two Faces of AI: Mind Replicant and Mind Cravings

Let us start with how we provide AI with the data it requires to replicate us.

AI as daily support

Every day, we use many different apps in our personal life which provide AI with our individual data. While AI captures this information to feed datasets, it also uses it as amalgamated content to further enrich its queries.

In our professional life, every trade is now equipped with basic administration tools like Microsoft 365 or equivalent. It is also the case at the trade level. For instance, with change management, a trade based on a person’s capacity to accept organisational changes, there are many AI tools that are available to bring clear benefits to organisations as they:

• Save time by handling repetitive tasks

• Reduce cost by automating processes

• Improve communication to help your team stay informed

• Boost employee engagement by collecting and acting on real-time feedback

• Streamline training & onboarding by simplifying team learning

• Minimize risk by flagging problems early.

There are methodologies that offer ‘systemic approaches’ like PROSCI with an AI component.

More complicated organisational changes can still benefit from these methods and tools but will find them limited to address the complexity of the constant changes occurring at the internal and external stakeholders' levels. Such transformations no longer use a waterfall project management approach but rely on Agile-like methods and change management strategies and tools on steroids (Design thinking, pilot end-user testing, etc.).

SelFou, in the context of managing a person’s aptitude to change and capacity to adapt, reached out to Peter Magnani, CEO and Founder of Beam Resilience Intelligence System.

We first spoke about the company’s choice of word ‘resilience’ over wellbeing and wellness. Peter explained: “It is a question of scope. Beam is a “self-learning” technology empowering the individual to minimise micro-stress and improve personal resilience and wellbeing through holistic, actionable insights and individualised support. Beam is less about mimicking consciousness and more about providing practical, preventative, and ethical support to individuals and organisations.”

Through this AI-powered data-driven solution, Peter opened a new window of support available to the human side of complex (or simple) changes/transformations. “For organisations, Peter continues, this means the delivery of aggregated, anonymised intelligence is far more accurate and actionable than questionnaires or annual surveys, because it’s based on passive behavioural and emotional data in real time. This dual perspective makes Beam uniquely positioned to support both the individual and the organisation throughout constant cycles of transformation.”

SelFou also contacted behavioural sciences firm, BIT (the Behavioural Insights Team) which is a global research and innovation consultancy which combines a deep understanding of human behaviour with evidence-led problem-solving to improve people’s lives. Among other events, BIT offers training on the psychology of organisational change and how to master organisational dynamics, which equips attendees with practical, low-cost behavioural science tools that tap into the drivers of human behaviour and significantly increase the likelihood of achieving one's desired results.

Lastly SelFou, for this blog, explored with Amanda Kirby, the effect of neurosciences and neurodivergence on organisations. Amanda founded Do-IT Solutions (doitprofile.com), an online neurodiversity screening and assessment platform designed to help organizations and individuals identify strengths and challenges associated with various neurodivergent conditions, such as ADHD, dyslexia, autism, dyscalculia, and dyspraxia. Her input on ToM was invaluable.

We have explored where the data comes from. Now we dive into how it is modeled.

Going behind the scenes—how AI replicates us

When we did our research for this blog, we wanted to understand how scientists empowered AI to be able to ‘think’ and be like humans. As Amanda introduced us to the ToM, Michal illustrated how it is related to language and Large Language Models (LLM).

We then wondered if, along with behavioral sciences, BDIs (beliefs, desires and intentions), resilience and other components, we could provide a picture of how AI aspires to become more human than humans. Here is what we crafted for you. For a fulsome image, you may want to dive into the sources themselves.

ToM

‘This ability—typically referred to as “theory of mind” (ToM)—is considered central to human social interactions, communication, empathy, self-consciousness, moral judgment, and even religious beliefs. It develops early in human life and is so critical that its dysfunctions characterize a multitude of psychiatric disorders, including autism, bipolar disorder, schizophrenia, and psychopathy. Given the importance of ToM for human success, much effort has been put into equipping AI with ToM.’ Source" idem as for the visual of the BDI PSD above https://www.pnas.org/doi/10.1073/pnas.2405460121- note that all references are available in the original text.

LLM

‘Large language models, or LLMs, are advanced artificial intelligence systems designed to process and generate human-like text. They achieve this by analyzing patterns in vast datasets containing language from books, websites, and other sources. These models predict the next word or phrase in a sequence based on the context provided, allowing them to craft coherent and contextually appropriate responses. Underlying their functionality is a neural network architecture known as a “transformer,” which uses mechanisms like attention to identify relationships between words and phrases.’ Source: (https://arxiv.org/html/2501.15355v1)

More please

‘Humans automatically and effortlessly track others’ unobservable mental states, such as their knowledge, intentions, beliefs, and desires. This ability—typically called “theory of mind” (ToM)—is fundamental to human social interactions, communication, empathy, consciousness, moral judgment, and religious beliefs. Our results show that recent large language models (LLMs) can solve false-belief tasks, typically used to evaluate ToM in humans. Regardless of how we interpret these outcomes, they signify the advent of more powerful and socially skilled AI—with profound positive and negative implications.’ Source: Michal Kosinski, Evaluating large language models in theory of the mind tasks, PNAS, 2024, Vol. 121 No. 45, introduction: Significance.

ToM with a cherry on top: ToM-agents

For those of you who are interested, this study extended the ToM to open-domain conversational interaction scenarios by proposing generative agents leveraging a novel BDI tracking paradigm: ‘ In contrast to existing models or paradigms that limit task-specific computable ToM models to binary conditions (true belief or false belief), the proposed ToM-Agent can disentangle the belief and confidence based on psychological research. This allows for the simulation of generative agents engaged in open-domain conversational interactions that consider varying degrees of confidence in different mental states, such as BDIs, knowledge, and more.’ Source: (https://arxiv.org/html/2501.15355v1)

As always, dear Selfies, we hope this has brought you some insight or inspiration to further your own organisation or personal understanding.