Artificial intelligence (AI) is a controversial topic. There’s a lot of debate about what it’s capable of doing, how far we can advance it, and if we should be developing it at all. While it’s a bit too late to talk about whether or not we should be advancing AI, there’s room for discussion about which direction to go in with its app development cost.
The main goal of AI is to make life easier for humanity. It’s a tool that will help us make many aspects of our lives more efficient, safer, and more fulfilling. However, if we want to integrate AI seamlessly into life, we need the systems to operate in a way that’s simple for humans to understand and control.
There are numerous examples of AI failures that are both amusing and disturbing. If we want to avoid future failure in AI systems we may one day come to rely on, we need to rethink the way we teach machines to think, understand, and interpret.
Traditional Logic in AI
AI is programmed to think logically. Systems that use AI draw from the available data to deduce or induce logical conclusions. AI that’s equipped with powerful machine learning capabilities can get really good at coming up with logical conclusions if you give it enough data. However, it depends heavily on human input to create parameters around the intended results of analysis.
The way we program AI systems to think determines not only how they will act, but also how well those actions align to the real goals we had in mind. It’s remarkably difficult to align an AI system’s goals to your own in an exact enough way to prevent the system from ever deviating.
For example, if you programmed an autonomous, self-driving car to get you home as quickly as possible, you could end up with wildly different results depending on what other goals the AI already has. If the car had no other existing goals, it could become a danger to other drivers by going at 150 mph down residential roads and swerving to avoid other vehicles or pedestrians. You might show up in bad shape, if you made it home at all.
While this is a silly example, it does help show why logical thinking alone may not be the answer to programming AI. Driving as fast as possible is the logical way to get somewhere quickly, but it doesn’t take into account how the passenger and others around the vehicle might be affected by this course of action.
What is Affective Logic?
Affective logic, sometimes called emotional logic, refers to the idea that feelings and emotions are separate from other thinking processes. This theory says that feelings and emotions use different cognitive abilities that regular logical thought processes, that they are independent of one another.
In AI, affective logic is sometimes synonymous with terms like artificial empathy or emotional intelligence. While existing AI is able to make deductions and inductions to come up with a logical conclusion, this process has its limits.
Why Affective Logic in AI Matters
AI is notoriously bad at interpreting highly complex situations, especially those involving human emotions or feelings.
Affective logic is helpful for machines interacting with humans. Not only does it allow the machine to make decisions that are more in line with what a human would do, but it also allows humans to understand the decision-making process and what the AI system is doing.
People are better able to interact with AI if the AI can communicate in a way that’s familiar and comfortable. Because the nature of emotion and feeling is not as straightforward to teach as basic logic, it’s tricky for an AI to be trained to respond the right way. The AI needs to be able to recognize a sentiment and tailor responses to fit it, often needing to mirror emotions in order to elicit the best response from the person interacting with the AI.
Often the way we say things is just as important as what’s being said. If AI can be trained to use affective logic along with existing thought processes, they will become more responsive to human needs, allowing us to integrate AI more seamlessly into the tech we use every day.
The purpose of AI is not to replace humans. Some of the best received AI tech today is ostentatiously un-human. Voice assistants used by Amazon, Google, and Apple are all examples of AI systems built to assist humans without trying to pass off as a human. Everyone knows they’re not talking to a real person, but if the system does what it’s asked to do, we generally accept it as good and useful.
Voice assistants have been vastly improved over the years. As certain machine learning concepts have been used to improve functions like speech recognition and responses, the assistants have become more useful. If you use your voice assistant often enough, it can learn to recognize the way you speak so that it will hear you and respond more accurately each time.
Potential Business Applications for AI with Affective Logic
AI that can use affective logic would be better suited for customer service than existing iterations. There are already numerous examples of AI being put to use for customer service, helping companies sort out piles of customer data or find solutions to problems. However, if AI could be better equipped to understand and respond using emotional intelligence, it could be used to play a more direct role in interacting with customers.
One of the biggest frustrations people have about automated customer service, such as AI-powered chatbots, is that they just don’t understand what you need as well as a human service agent would. Responses are overly mechanical and may not address the actual issue in a way that satisfies the customer.
A 2019 survey showed that 80% of people polled prefer to interact with human customer service agents rather than electronic systems. Preferences varied based on the industry and purpose of the interaction, but for most customer service applications, there was a preference for human contact.
If AI were more effective at understanding and appropriately responding to human needs, including responding to the customer’s emotional state when they contact the company, people would likely have a better experience with automated customer service.
The biggest complaint about AI customer service is that humans understand a person’s needs better than electronic systems. Affective logic will help AI better understand the scope of human needs in order to address them quickly and efficiently.
AI chatbots are not the only way to apply this tech to customer service. Some of the best IVR (interactive voice response) systems could be made even better with AI that could interpret and respond to human feelings and emotions. With the addition of effective logic AI systems, all non-physical customer service exchanges could be automated without sacrificing the quality of service. Also, there are many other sectors besides customer service that could be highly beneficial if AI had a more affective logic. For instance, studies and research suggest that the recruitment industry could change forever with AI. Yet, it is still not fully implemented because recruiters not only analyze resumes but also take time to analyze other aspects of in-depth-for.
Is Affective Logic in AI Possible?
The short answer is yes, it is possible to program AI to use affective logic. In fact, many existing systems utilize some form of emotional intelligence to understand and manipulate human desires and behaviors. Advanced algorithms developed by social media networks, search engines, email sending platforms, and online ad agencies have emotional intelligence programmed into their AI.
This existing programming is responsible for interpreting and predicting human emotions based on data from a person’s online presence. The AI uses machine learning to draw conclusions about human actions and reactions based on the data it reads about previous related situations.
These algorithms are good at what they do. So good that sometimes the people that created them don’t even know how the AI came to a specific conclusion. The algorithms have gotten so advanced that they’re often better at detecting specific human emotions than people are. However, they still have limited application.
The next step is to teach AI to act on emotions in the same way humans do. This is a much trickier task. It’s somewhat straightforward for an AI to look at a large collection of data and learn to recognize emotion or sentiment in videos, pictures, and text. While it’s not an easy process, machine learning makes it possible with enough time and resources today.
AI is not human and will likely never be able to feel in the same way a person does. However, it may be possible to program emotional logic into an AI system so that it mixes feeling and emotion into the decision-making process. We need AI to be able to understand the emotional response elicited by a specific situation, then use that understanding to add weight to decision-making.
Creating a framework for affective logic is difficult. This is partially because emotion isn’t fully understood in humans yet. Because we don’t fully understand how emotion works in our own minds, it’s difficult to replicate it in coding. The other major difficulty is sorting out exactly which emotions matter and how to utilize them appropriately.
For example, hunger or fatigue are useless emotions for AI. These emotions are only useful when there is a physical body that needs to be maintained. However, empathy and compassion are qualities we would desire an AI to exercise in decision-making.
An AI programmed to assist doctors in patient treatment would be of little value if it was unable to take emotion into account. Medical decisions are often linked with both logical and emotional decision-making processes. To be useful in this function, the AI would need to go beyond simply recognizing emotions and actually integrate internal emotions into logical decisions.
As of yet, we have a long way to go for this particular tech. It’s theoretically possible, although we’re likely a long way off from creating functional affective logic in AI.
If we want to make an AI that utilizes something similar to human emotion for decision-making purposes, we need to make sure we account for certain challenges that go beyond the scope of the affective logic programming itself.
Advanced AI has certain drawbacks that should be addressed. As long as we enter into research and experimentation understanding the risks and challenges we may need to overcome, we can move in the right direction.
Privacy is one of the most obvious challenges. AI needs a lot of data to learn and draw the right conclusions. The problem is that this data may be sensitive to theft or misuse. Algorithms using AI often capture information about a user’s age, gender, race, sexual orientation, location, political views, religious affiliations, and more. User’s interactions with the algorithm are tracked and data is collected in a central system.
AI needs a lot of data in order to operate appropriately. Keeping this data safe and secure is important, but it’s just as vital to use data in a way that protects user privacy. This is an immediate problem we are already facing, but it can only grow as we develop more complex AI systems.
Any piece of tech can be misused for malicious purposes. There are already cases of people using advanced natural language AI programs to commit fraud. The more systems in which AI is integrated, the more opportunities there are for that AI to be accessed and changed.
AI itself cannot be inherently evil. It does what it’s programmed to do. If the goals given to the machine are either poorly formulated for evil in nature, the AI will act maliciously. We need to be careful that AI will only be used to benefit the world rather than to control or harm it.
This likely means developing security standards and AI ethics, then working on how to enforce them in a useful way. If we begin to focus on this problem now, we may be able to keep up with advances in the tech to create useful regulations before they’re needed.
AI built with affective logic is something we can almost certainly look forward to in the future. It may be awhile until we see it in action, but it’s best to start thinking ahead to look for ways we can use it to benefit the most people without causing harm. It has great applications in business, medicine, and so much more, as long as we take care to put the right foundational steps in place today.