Publication
Blog
Artificial intelligence (AI) is a way for computers to simulate human intelligence to perform complex reasoning and problem solving. But there are many ways AI can do this, and it is no stranger to taking inspiration from biology.
Like many other fields of science, inspiration for AI often comes from nature, and a whole range of nature-inspired AI exists, such as genetic algorithms and neural networks. For example:
Neural networks, in particular, are part of a ‘thinking human’ approach to developing AIs. However, thinking ‘like’ a human is more than just a system that processes data and makes decisions using data. Emotional intelligence also has a part to play.
Research in cognitive science suggests that emotions play a crucial role in decision making. They can shape decisions in various ways, such as influencing the content and depth of our thoughts and reaching specific goals. They can even act as shortcuts in some decision making.
Let’s consider an example. Say you are in a shop and see two identical t-shirts: one red and one green, both at the same price, quality and style. You choose the red top over the green one because it gives you a happy feeling. It may remind you of a happy memory, or it has a better vibe, or you were in a red kind of mood that day.
You then realise you need a second T-shirt and return to the shop a week later. But today, you choose green. Perhaps you want more variety for fear of being labelled ‘the red shirt guy’, or maybe your mood is a bit low today, and you decide to go for a more calming colour. Either way you then opt for that one today.
This concept is part of a broader understanding of how emotions, whether related to the current context (integral emotions) or not (incidental emotions), can significantly shape our decisions and interactions with others.
At their base level, neuroscience describes emotions as part of a complex biological system of neurobiology and neurochemistry, coupled with lived experiences and memories.
The book ‘Neurobiological Basis of Emotions’ states that “emotions are embedded in neural networks”, that's ‘biological’ neural networks.
The book talks about how emotions are interwoven relations of brain, body and memories. It goes on to explain the complex systems of chemicals and hormones like dopamine, a hormone associated with the brain’s reward system. It influences motivation, memory, and concentration or how serotonin, often referred to as the ‘happiness hormone’, promotes a sense of calm, and how injury to the brain can impact emotional processing.
The paper ‘Neurobiology of emotions: anatomy, neural circuits, and alexithymia’ argues that emotion can be understood through three interacting processes: affect generation, affective response representation, and conscious access.
Affect generation happens through the interaction of the front part of the brain and deeper brain areas, which are important for starting emotions. When we express our feelings, other parts of the brain, including those related to movements, are involved. On the other hand, being aware of our emotions involves different brain areas that help us think about and reflect on how we feel in a biological system.
Which AI does not have (so far).
The role of emotions in shaping our decisions and interactions with others is a key aspect of human behaviour. Both integral and incidental emotions play a significant role in influencing our thought processes and behaviours. Emotional intelligence, sometimes called EI or EQ (emotional quotient), is a practical skill we use daily to perceive emotions.
It’s about awareness of emotions in oneself and others and using this awareness to guide our thinking and behaviour. It can be broken down into the following points:
(Purushothaman, 2021, Emotional Intelligence, p34)
This is where it gets tricky. If industry’s goal is to achieve human intelligence, how can it simulate emotional intelligence?
There are approaches such as visual indicators with image recognition, such as a smile or frown for happiness or sadness. Sentiment analysis with words and tagging negative and positive phrases with natural language processing. These align with points 4 and 5 but lack the others. This is because they are representations of emotion, not emotions, such as lighting or composition, to express emotions in a visual medium.
Emotional intelligence covers more than just the recognition of emotions. It includes self-awareness and self-regulation of one’s feelings. To have self, one needs to have the capacity for emotions in some form.
Therefore, if our opening statement about ‘computers to simulate human intelligence’ is the goal with the likes of general AI (AGI), current technology and AI on the market today have limitations in replicating emotional intelligence like a human.
An argument could be made that the training data is a substitute for lived experiences. But as previously mentioned, emotions are rooted in the chemical, biological and management of one’s self. It’s how two people with the same educational background and resources can approach the same problem but come to different solutions or even the same person at other times if we go back to our t-shirt dilemma.
This leads us to a question of AI. Is it to be a tool that improves tasks or to imitate humans, emotions and all? If it is the latter, AI may need to go down the biological route more and introduce all the weaknesses and strengths that come from its complex chemical makeup of emotions.
We are a long way off biological machines in the current state of technology, which leaves us with the challenge of further advancing our current mechanical and digital technologies.
It's important to realise this limitation and differences in AI to human intelligence and propose considering it more a tool that can simulate these features if needed.
This has positive aspects for AI, such as removing potential bias through logical thinking and complex data. However, not letting emotions dictate decision making could still result in bias.
Biases can result from statistical under-representation in the training data, resulting in underrepresentation. This bias can significantly negatively affect the outcomes of decisions. For instance, in a study by Lauren Rhue involving the emotion detection software Face++ to analyse emotional responses, it was found that it displayed racial disparities in their emotional scores. Tending to assign negative emotions more frequently to the faces of black men.
At the HCI 23 conference, experts in the field of Human-Computer Interaction delved into the discussion of the ethical and social implications of emotional intelligence in general. The use of this technology in surveillance systems has drawn criticism for its perceived resemblance to a lie detector, as highlighted in Wakefield’s BBC article from 2021. There is a growing consensus about the necessity of establishing regulations to govern the use of emotional AI, recognising its potential impact on privacy, human rights, and societal well-being.
So, I asked an AI, in the form of Microsoft’s Co-pilot. Its response is below.
“Artificial Intelligence (AI) can be viewed from both perspectives. On one hand, AI is a tool that improves tasks. It can automate repetitive tasks, analyse large amounts of data quickly, and perform complex calculations that would be difficult for humans. AI can also assist in decision-making processes by providing insights based on data analysis. It’s used in various fields such as healthcare, finance, transportation and more to improve efficiency and effectiveness.
“On the other hand, some AI systems are designed to imitate human behaviour. This includes natural language processing (understanding and generating human language), computer vision (interpreting visual information), and machine learning (learning from experience). These systems aim to replicate human abilities to understand, learn, and respond in a way that appears human-like.
“However, it’s important to note that while AI can imitate certain aspects of human intelligence, it does not possess consciousness or emotions as humans do. AI operates based on algorithms and data and does not have personal experiences or subjective consciousness.
“In summary, AI is both a tool to improve tasks and a system to imitate human behaviour, depending on its application. But it’s not a replacement for human intelligence or experience. It’s a tool designed to augment human capabilities and make tasks easier and more efficient.”
We know there were no emotions mentioned in the creation of the above statement, and there was no mix of dopamine and serotonin at the small joy of completing a task.
No increase in cortisol from the anxiety of someone reading what you wrote (which I, the author, have been feeling a bit now writing this). Yet, it is performed as a task asked for by a human and at this stage is all we need for most applications – a tool.
So, it is important to note that, while AI can imitate certain aspects of human intelligence and in some cases surpass it, it does not possess emotions and emotional intelligence like humans do.
AI operates based on algorithms, probability and data and does not have personal experiences or subjective consciousness. With the advanced in the current market, it is easy to lose sight of this in the pursuit of true intelligence. So, as not to lose sight of the risk and bias that can arise, as intelligence is not just one thing and focus on what it can achieve now as a tool.
Reference:
Purushothaman, Rajagopalan. Emotional Intelligence, SAGE Publications India Pvt, Ltd., 2021. ProQuest Ebook Central
Opencast has been building new competency frameworks that offer all its people a clear understanding of the skills and knowledge they need to perform...
Loading...
