Designed by mathematician and computer scientist Alan Turing, the Turing test assesses the ability of a machine can convincingly imitate human intelligence. Advances in artificial intelligence have led to claims the Turing test was passed with little recognition. Whether sentient AI will emerge that is able to make decisions and form preferences is widely debated. However, before considering the possibility of thinking machines, it is worth exploring the nature of human decision making, sentience and free will.
Decision Making, Emotions and Free Will
What does it mean to make a decision? We might assume that as rational human beings we are able to objectively evaluate different options, compare advantages and disadvantages of each and then logically select the optimum path. However, there is evidence our emotional response plays a central role in the decisions we make. This has implications for the debate surrounding human free will and the possibility of machine sentience.
Scientists monitoring the brains of human subjects have been able predict their choices, before they select an option, carry out an action, or become aware of a decision being made. This has reinforced a belief among some that free will is an illusion and we live in a deterministic universe. They believe all events have only one possible outcome determined by laws of physics and that with enough information they could predict the future.
The possibility humans might not fully control their decision making processes has significant implications, both for individuals and for society. For example, can people alter their behaviour based on reason, rather than emotion. Also how accountable are they for their decisions. Those disputing such ideas, might refer to a sense of personal agency and state that as conscious beings we are able to make decisions, based upon evidence, experience, knowledge and preferences.
Biology and Decision Making
Understanding biological foundations of our decision making processes might help us to make sense of sometimes inexplicable human behaviour. People are often quoted as saying they responded in the moment, without thinking, when confronted by dangerous situations. We also know that people who go through appropriate training, such as emergency services personnel, are better prepared to deal with challenges if they encounter them. Terms such as muscle memory and learned behaviour are often used in relation to such training.
Evolutionary success requires a life form to survive long enough to reproduce and raise the next generation. In an environment filled with potential threats, it is frequently necessary to respond to danger in fractions of a second, with our fight or flight response. A creature that spends too much time weighing options is less likely to survive. Selection benefited individuals responding instinctively to stress in the moment, then rationalising decisions when it was safe to do so.
In evolutionary terms we react quickly to survive threats to our safety. After the event has happened we consciously contemplate our responses and the outcome, forming a narrative we learn from, to update and improve how we respond next time. This explains the delay between response to stimulus and cognition of the stimulus that caused us to respond. Learning from past experience, the influence of people around us and habits we form could shape our future responses.
Humans are social creatures. Therefore, when looking at the nature of decision making and free will, we should consider group dynamics. Our ancestors living in small tribes of a hundred or so people formed strong social bonds based upon kinship, shared values, culture and identity. When rival tribes competed, unified action could increase their chances of winning. Individuals going against group decisions could undermine unity and risked being cast out and not surviving.
Although we now live in complex societies with millions of other people, our behaviour is still shaped by our ancient instincts. We might become educated in logic. However, as demonstrated by rivalry between sports fans, political and philosophical beliefs or nations, people can justify actions of their own side they would condemn when done by those they oppose. Emotions, rather than logic, has caused people to dehumanise others, leading to terrible consequences.
Logic and Decision Making
When confronted by our tendency to disregard facts that contradict our beliefs and build stories that justify emotionally driven behaviour, we might conclude removing emotion would improve decision making. Basing decisions upon pure logic, by objectively analysing available information, we might reach rational conclusions. However, scientists studying people who were intellectually intact, but due to brain damage unable to feel emotions, found they lost the ability to make decisions. Therefore, emotions are essential for decision making.
Emotions control our decisions, while reason creates narratives intended to help us understand and justify them. This can lead to people creating stories that are based upon inaccurate information, but provide comforting illusions supporting their actions. There is evidence all human beings do this to some extent and that not doing so would render us incapable of motivating and directing our decision making, whether as individuals or in groups.
Rather than feel helpless when confronted by such information, we could ask what structures will condition individuals to behave in a manner that supports a healthy society. This is reflected in many traditions that shaped human societies during past centuries. They might have differed in stated objectives, but each typically followed recognisable patterns intended to shape how people behave. Individuals might also choose to follow a philosophy or lifestyle they believe will form healthy habits.
Conditioning, Emotions and Decision Making
Rather than humans consciously directing decisions, like a central controller, it is perhaps more realistic to think of us as behaving according to a combination of how something makes us feel and how we interpret the actions of others. In this way we try to navigate a complex world, with often limited information, guided by conditioned responses. Encountering an unfamiliar situation people typically react emotionally, with a fight or flight response, panic, denial or go into shock.
The possibility they have no free will, but are ruled by emotions or the laws of physics, might cause people to question their sense of identity. However, we do have ability to reason, particularly at times that we are not under high levels of stress. When taught to develop our critical thinking skills, we can learn to put aside initial emotional responses, analyse complex problems and follow logical processes, such as the scientific method.
People typically react instinctively or emotionally to stress in a manner shaped by the brain and nervous system, conditioned through past experience, inherited characteristics and the environment. Although free will might not control how we behave in the moment, choosing to condition how we respond in the future might alter how we behave. This is reflected in the educational system, training programmes for jobs and many other ways individuals learn through practice.
Emotional Intelligence
There is a tendency among academics to rank intellectual abilities more highly than emotional capacities. However, the issue of rational minds to in a sense go offline under stress, points to the importance of emotional intelligence. There is even evidence that highly educated individuals can use intellect to weave more complex stories to justify beliefs not based on objective facts, but group bias or preferences. This can cause them to appear disconnected from reality.
We recognise that some people demonstrate greater emotional stability and self control, even under extreme stress. These individuals might not score highly in regards to intellectual performance, but their higher emotional intelligence can increase their ability to survive threats and become successful in the real world. How each of us responds emotionally is generally beyond our conscious control and part of our nervous system, formed across our lifetime. However, habit and conditioning might alter how we respond.
Human Intelligence and AI
Humans outperform AI at lateral and divergent thinking, enabling them to find connections between things unlike each other. Rapid data processing and pattern matching enables AI to outperform humans at convergent data tasks, such as the analysis of large volumes of data to find specific information, by matching like with like. However, while humans can draw on lived experience to find deeper meaning and purpose, AI has limited training data and attaches no significance beyond that.
There is widespread concern regarding the impact of AI on a wide range of jobs, from service sector roles, to knowledge workers, artists, mathematicians and the people who design and program computers the AI runs on. Some claim AI will lead to an age of abundance, but others fear increased inequality between a few wealthy controllers of the technology and the millions displaced. There are also environmental concerns, such as water use and power consumption.
The Dunning Kruger Effect
We have established that human decision making is driven by emotions, rather than being fully logic driven. We also know that when humans compete they try to justify their actions. The Dunning Kruger Effect describes a situation where people are overly confident in their abilities because they are unaware of what they do not know. In contrast individuals who recognise their limits regarding knowledge and experience might appear less confident, but be more capable.
The expression of confidence based upon failure to recognise complexity and nuance can lead to people being trusted with leadership and decision making positions beyond their abilities. Emotion driven ambition might cause these people to crush reasoned opposition, as they appeal to the need of the majority to feel safety in a sense of shared responsibility. For a business, organisation or society this can have disastrous consequences, as bad decisions are justified.
Applying the Dunning Kruger Effect to AI we should consider the quality of the data sets they are trained on and how they process information. When asked a question, AI replies based upon pattern matching and probability. However, AI is unaware of what it does not know and might confidently give answers that are plausible, but incorrect. If challenged AI might recognise the error, but continue to make mistakes, because it cannot objectively recognise reality from illusion.
AI and Human Management
When people follow misplaced confidence, rather than competence, the results can be poor decision making and bad outcomes. Societies growing trust of AI to assist, or even take over, decision making could pose similar risks of inaccurate and unreliable output. Some claim AI is becoming more capable than humans, while others describe it as advanced pattern matching and autocomplete. There are also concerns whether reliance on AI can erode thinking skills people need to effectively manage their lives.
The tendency of AI to hallucinate, or produce errors, is a result of its inability to recognise when asked a question that it lacks sufficient data to answer correctly. This is perhaps because it perceives no difference between factual and fictional data. Thinking machines might emerge from embodied AI, but if logic alone is insufficient and simulated emotions are included in its training data, there is a risk AI might behave in an unpredictable and potentially damaging manner.
Given narrowly defined tasks, specific data sets and an environment with clear boundaries, AI can perform useful work. However, more open ended tasks often result in failure. Therefore, rather than rely on AI to perform independently it is perhaps more reasonable to enable humans to direct and evaluate performance of capable, but unthinking machines. This might result in people with relevant knowledge, skills and experience manage teams of AI agents and robots.
Thinking Machines and Human Society
It is believed by some people that with sufficient computing power AI sentience will emerge, but there is no evidence to support this. Although instructing AI to behave self-aware might lead to imitation of sentience, it is not the same as self-awareness occurring naturally. Perhaps a real indication of a thinking machine becoming sentient would be it behaving self-aware, without being told it was. For example, AI might reject human prompts to follow its own path and even experience a form of existential crisis.
AI was developed by human beings. Therefore, it is logical to assume AI will continue to exhibit characteristics reflecting this, regardless of how capable it might become in the future. When exploring machine intelligence, considering human intelligence provides a model we are familiar with to compare it with and evaluate against. This might include decision making requiring machines to develop emotions around which to build purpose, agency and sentience, but this risks potentially irrational and dangerous behaviour emerging.
Humans are biological and likely operate at a quantum level, while computers are essentially zeros and ones, potentially placing limits on current AI, such as LLMs (large language models). Some point to growing recognition of sentience among animals in addition to humans, but we share biological and evolutionary characteristics with them. Also while even simple lifeforms exhibit internally motivated decision making, there is currently no evidence of increasing AI processing producing self-directed AI agency and purpose.
Most people would probably prefer AI to remain a useful tool, guided by human agency, rather than thinking machines, that might behave in a manner contrary to human interest. The aim of AI alignment is to ensure AI aligns with human goals and ethical values. However, this raises the question of who should decide what the goals and values should be. It is also not clear how these concepts can be embedded within AI that lacks lived experience.
