Bots are dead, long live AI (Part 2/5)

How AI becomes intelligent, and what it means to be intelligent

Salovey and Mayer found different interpretations for human intelligence throughout history. One of the first definitions by Descartes supposed that intelligence is the ability to judge true from false. Later on, Daniel Wechsler’s idea of intelligence is the aggregate or global capacity of the individual to act purposefully, to think rationally, and to deal efficiently with his environment. Wechslers definition includes the distinction among mechanical, abstract and social intelligence. However, research has kept on searching for a set of dimensions that make up intelligence. A turning point in this research was in 1983, when Howard Gardner published his book Frames of Mind. His argument that intelligence does not root in one single trait was challenging the renowned testing of intelligent quotients in education. Arguing that there is no single right answer to certain questions, Gardner came up with a model of multiple intelligences. Over time, the model of multiple intelligences has been refined and enhanced, which brought Karl Albrecht to distinguish and reframe six multiple smarts:

• Abstract Intelligence: symbolic reasoning
• Practical Intelligence: getting things done
• Emotional Intelligence: self-awareness and self-management
• Aesthetic Intelligence: the sense of form, design, music, art and literature
• Kinesthetic Intelligence: whole-body skills, dancing or flying a jet fighter
• Social Intelligence: dealing with people

With the six-dimensional model of intelligence in place, Albrecht argues that the combination of the listed intelligences and the synergies resulting from the combination form a portrait of a true Renaissance person. Again referring to Alan Turing, the motivation is to apply the concept of intelligence to create artificial intelligence in computers to form a true Renaissance agent that is capable of behaving like a human.

Intelligent Agents
Alan Turing proposed the imitation game with the question: Can machines think? As a starting point for the field of AI, the imitation game is of special importance. Following Turings initial idea, artificial intelligence seeks to develop human-like intelligent systems. Next to the main domains of science and engineering, AI is also inspired by other disciplines like mathematics, linguistics, psychology, neuroscience, statistics or economics. Thereby, characteristics of intelligence such as adaption and learning as well as acting in an environment, problem solving and planning, perception, or natural language processing make up the primary scientific goals. The term artificial intelligence was coined by John McCarthy and AI pioneers like Herbert Simon, Marvin Minsky and Allen Newell among others, during a summer workshop at Dartmouth in 1956. In 2015, Mikolov, Joulin and Baroni state that the time for intelligent machines is ripe. Sufficient computational power, immense amounts of data, plus complex machine-learning methods allow for the creation of general purpose, sophisticated intelligent systems. The focus is on intelligence in terms of communication and learning.

Tecuci states that the purpose of artificial intelligence is to create intelligent agents which are capable of achieving goals by knowing their environment and the stakeholders they have to deal with, as well as memorizing the gained information and improving their behaviour through learning. AI research has focused on creating these kind of agents in the past. Combining two types of agents, autonomous interface agents give the possibility for both the user and the agent to interact in parallel through an interface. As Lieberman puts it: […] it follows that there must be some part of the interface that the agent must operate in an autonomous fashion. The user must be able to directly observe autonomous actions of the agent and the agent must be able to observe actions taken autonomously by the user in the interface.

Autonomous, intelligent, interface agents inhabit social media ecosystems and bring up challenges and new dimensions by emulating content, activity patterns and sentiment expressions. Social media bots autonomously produce content and interact on social media to influence humans. This challenge is more evident than ever in the context of social media, due to limited attention and technological constraints which are exhausting the power of humans. Nonetheless, Tecuci deems AI as important due to the psychological and physical limitations of humans, such as limited attention span, memory or affection by stress. Social bots can be both harmful and helpful by aggregating content from different sources acting as a news feed or automatically responding to inquiries for brands and companies in customer care through interfaces. The agents create enormous amounts of content without delivering veracity of the information and the promotion of true information.

Therefore, Ferrara et al. identified various effects on humans that bots can create. For example, the bot effect explains that social botnets expose private data using vulnerabilities of social media users, or computational agents in- filtrating unaware populations and affecting their perception of reality. However, one of the biggest challenges is understanding just how harmful bots can become. Intelligent agents are able to search the Internet for information and media to create profiles, collect material and present it at fixed times as well as emulating content production and consumption. In addition, intelligent agents can post, comment, moderate and share content or answer questions. These actions can infiltrate discussions, connect to influential people and capture attention of many using natural language algorithms for the communication. Therefore, the boundary between human- like and bot-like gets fuzzier.

Socially Intelligent Agents
As described above, humans interact with intelligent messaging agents during work and private life. Unnoticed by the user, the software agent mimics human traits. Since the creation of intelligent systems like chat bots, code developed to have a conversation, the Turing test tries to differentiate human behaviour from the behaviour of computer algorithms. With the rise of a messaging ecosystem where messaging is the new platform and bots inhabit the place of apps, intelligent messaging agents raise the bar of the challenge, as they introduce new dimensions to emulate in addition to content, including the social network, temporal activity, diffusion patterns and sentiment expression. Although, coming back to the purpose of the test envisioned by Alan Turing, it is to determine how human-like computer systems behave. This includes the social capabilities of the algorithm. As Persson et al. put it: The ultimate purpose with socially intelligent agent (SIA) technology is not to simulate social intelligence per se, but to let an agent give an impression of social intelligence.

Karl Albrecht defines key aspects for the social intelligence of human beings, which computer scientist seek to implement into algorithms up to a certain point. Albrechts findings describe the key principles of clarity, situational awareness, empathy, presence, and authenticity. Therefore, situational awareness is described as the ability of understanding the situation and its circumstances creating a social radar to detect what is going on around a situation. Therefore, studies towards situational and context awareness of intelligent agents are performed by various teams. Context awareness al- lows applications to adapt themselves to their computing environment in order to better suit the needs of the user. The topic of contextual awareness also finds application in general artificial intelligence (AI) research like swarm robotics. Persson et al. mention that socially intelligent technology needs to understand its own presence and the context around itself, to further take other agents or humans into consideration. Further, Persson et al. take advantage of primitive psychology and life-preserving aspects of needs, desires, sensations and pain to construct socially intelligent agent that can show empathy. Since, the representations include ethical and moral assumptions, socially intelligent agents have to be designed with great care to prohibit discriminatory acting or thinking and therefore, communicate their behaviour in a clear manner. However, a gap in literature and research regarding the social intelligence of intelligent agents can be identified going along the line with social intelligence capabilities of authenticity.

____
This post is the second in a series of five extracts from my Master’s thesis revolving around socially intelligent artificial intelligence. Thanks for reading! :) If you enjoyed this article, hit that ❤ button below. Would mean a lot to me and it helps other people see the story.