1. Introduction: The Mission of the Learning Ecology

2. The Challenge of Soft Computing Intelligence

3. Fuzziness as Stimulus for Learning

4. Learning for Decision Support versus Learning for Decision Creation

5. Learning to Understand

6. Learning to Create

7. Conclusion

References

1. Introduction: The Mission of the Learning Ecology

Ecology studies the web of dynamic interactions of the living creatures, including humans, and their environment - natural and artificial (human-made). Learning is a process which is vital for sustaining the integrity of this web and hence for sustaining the life and its unfolding. 

Learning ecology focuses on factors and conditions facilitating the process of learning and searches for ways to increase its efficiency, in the sense of opening new possibilities for realization of the self-organizing impetus of the living entities, at any level of the web of interactions. 

Plants and animals learn to adapt to the changing environment in order to survive, reproduce and increase their fitness. Some animals easily learn to follow human instructions and develop behavioral patterns classified by people as "clever", "friendly", "faithful", etc. 

Similarly to other species, people learn how to better cope with the changes in their environments. Some people learn to be aware of the events of their experiences and to make sense of them not as isolated events but in connection with each other; for others, learning constitutes the meaning of their lives.

Learning ecology considers the process of learning as essentially holistic - not only the human mind - reason, logic, ability to think and decide - is the most important agent in this process. Equally important are also the human heart - feelings, emotions, ability to love and care - and the human soul - intuition, inspiration, ability to aspire and meditate. The heart and soul factors are vital for manifestation of human creativity, and without creativity learning is a mere repetition of knowledge borrowed from books and gurus. 

Intelligent machines learn to do things, which are hard or impossible for people, like processing large files of data and recognizing patterns in them, solving problems with high computational complexity, moving and working in environments dangerous for human life, etc. Like people, intelligent machines can learn from teachers (supervised learning) or from their 'own experience' (unsupervised learning) or by using the both types of learning in a process called hybrid training. The latter is widely used in soft-computing based on advanced fuzzy neural and fuzzy-genetic techniques. 

At the level of human interactions with nature, learning ecology explores and promotes people's holistic learning to sustain continuity of life on the planet and support its inherent urge towards integrity, self-renewal and evolution. 

At the level of human interactions with intelligent machines, learning ecology explores and promotes machines' learning to understand and 'compute' with human perceptions. It is this kind of learning that is at the leading edge of the soft computing (SC) in the beginning of the new millennium [1]. 

Why learning ecology is interested in promoting this kind of 'cyber'-learning and not just learning how to support human decisions? 
 
 

top | menu

2. The Challenge of the Soft Computing Intelligence

When SC-intelligence learns to interpret and work with human perceptions, it simultaneously learns how to 'experience' the world of people. People's perceptions are holistic - they are not products only of their minds; the whole complex of body, mind and soul participates in experiencing, and hence in perceiving, human reality, both inner and outer. 

Unfortunately, the rules of logic which our reason tries to follow since the earliest years in school, have caused obstacles for the perceptions to express their holistic nature. Rigid mental patterns, thinking stereotypes and standards, prejudices and all kinds of socially implanted world-views and concepts, often with hard-to-surpass boundaries, strongly impede our ability to perceive the world synthetically - as a wholeness in which all the phenomena and processes, be they natural or human-created, are inseparably connected. The analytical approach of science aims at dividing and separating, analyzing and classifying, defining and labeling. This approach works logically and efficiently in the artificial world of technological and engineering realizations, but affect illogically and disastrously the world of nature any time when trying to subsume its spontaneity, creativity and inherent freedom under the dictate of the reason. The latter aims at harnessing our ability to perceive and experience the world holistically. Nature has endowed us with this ability, and it is pity to see how we are losing it under the pressure of logic and rationality. 

The message of this paper is straightforward: while learning how to comprehend and operate with human perceptions, SC-intelligence is able to learn to revive, amplify and make work their holistic nature, and not let it die under the sole pressure of the human reasoning. The ability of SC-intelligence to do this is backed by: 

• its colossal potential for integrating and synthetically processing almost limitless amount of information, and 
• lack of pre-imposed stereotypes, rigid worldviews and prejudices, which are inevitable for the way people think. 

3. Fuzziness as Stimulus for Learning

The human perceptions are fuzzy - they reflect the fuzziness of what we know about the complexly interwoven, constantly changing and therefore difficult-to-predict dynamics of life and nature. Life and nature, even in their apparently simple forms of manifestation, are beyond precise definitions and descriptions. Even a human-created system, which is built of many elements complexly related to each other and open to unpredictable changes, escapes relevant precise descriptions (Zadeh's Principle of Incompatibility [2]). 

The fuzziness of our perceptions does not impede the natural urge to explore reality and make sense of what we experience; on the contrary, the fuzziness acts as a strong stimulus for learning and keeping awake our awareness about the changes that constantly emerge [3]. 

The more efficient our learning to interpret and work with the fuzziness of what we perceive, the more able we are to facilitate the emergence of autonomous decisions, that is, decisions born out of our own creativity and not just copies of already made decisions. While able to facilitate such acts of creativity, we minimize the need to repeat, follow blindly or only support what others say and do. 

The same is true for the SC-based intelligence. Once it begins to understand and operate with the fuzziness of human perceptions, it becomes able to learn how to do something more than just supporting the decisions made by a 'rational' expert or operator. With the help of teachers, it can develop a capacity to recognize and display the emergence of its own autonomous decisions. 

Here lays the crucial difference between a DEcision-Support Intelligence (desi) and an Autonomous DEision-Maker (adem). 
 
 

top | menu

4. Learning for Decision Support versus Learning for Decision Creation

Desi aims only at assisting human decision-makers. There is always an expert or experts who formulates (in a precise or in a fuzzy way) a specific problem, its restrictions and goals, and selects a method (or methods) for solving it and hence reaching (or approaching 'closely enough') the stated goals. The computing power of desi and the self-organizing capacity of its fuzzy-neural or/and fuzzy- genetic software are used to search for the optimal (or a 'more or less' satisfactory) solution and thus to support the expert's logic. Desi is of help when used for solving engineering problems related to the design and implementation of intelligent control systems. Desi is strictly rational and fully obeys the expert's rules of logic, usually tuned to take economically efficient decisions. 

If the expert makes a mistake, it is supported and often amplified by desi. Thousands of technological decisions, which turned to be disastrous from ecological point of view, have been conceived in experts' minds and supported by desi's computing power. 

Different is the role of adem; it is assisted by people, engaged in a specific fields of social activity, to learn how to understand their fuzzy perceptions. By recursively operating with these perceptions - comparing, combining, intersecting, juxtaposing, classifying, seeking for similarities, recognizing patterns, building clusters, etc., adem's neural or/and genetic network becomes 'intelligent' enough to interpret the meanings which they express. 
 
 

top | menu

5. Learning to Understand

This first stage of adem's learning applies fuzzy-logic-based techniques developed for computing with words and expressions of natural language by translating them into words and expressions belonging to the so-called 'generalized constraint languages' [3]. 

The meaning which adem extracts from an individual's perception must be as closed as possible to the meaning which this individual has inserted in his/her perception. The high degree of closeness is guaranteed through a recursive procedure that represents a kind of dialogue between adem and humans - a dialogue, which serves to clarify the meanings encapsulated in their perceptions. 

Adem's learning at the first stage is facilitated by the fact that we (and the language we use) are social products; the largest part of us repeats in words and realizes in actions 'truths' adopted or established by the society in which we live. We share similar experiential findings and borrow knowledge from similar sources - books, papers, talks, mass media and the world wide web. Therefore, there is a tendency towards similarity in the ways we perceive our world. This tendency is emphasized by widely spread and constantly repeated advises how the 'normal' people in a 'normal' democratic society should 'normally' behave and 'normally' perceive the changes that 'normally' occur in their lives. 'Normal' are people whose 'normal' behaviour is to make money and try to reach a higher social status or public estimation - these are considered as the most significant achievements in our 'normal' society - achievements which we are highly recommended to accomplish before we die (as if we need them in our next lives). Today's understanding of a 'normal' democratic society is a society, which is consumption-oriented, run by the money and the global power of the world's richest financial corporations and their visible or invisible bosses, and persistently brainwashed by the media to keep the poor majority silent and not rebellious, and to let them think that the existing forms of democracy provide all with equal opportunities to live and realize their potentials. 

The learning ecology considers this kind of social 'normalization', actively promoted through an aggressive corporative expansion, as a serious threat for learning of both the human and the machine intelligence. The introduction of standards and models for 'normal' behaviour in the social realities affects the meanings encapsulated in human perceptions of these realities by making them converge to rigid socially adopted meanings. Although such kind of meanings may technically facilitate adem's understanding of human perceptions, it could be fatal for human creativity and also for creativity of the machine intelligence, in the light of their future symbiosis. 

It is the unique richness, spontaneity and integrity of each individual's experience that support the holistic nature of one's perceptions, their creative impetus and authenticity. Therefore this richness needs to be preserved by society and not killed through adopted stereotypes, norms and global standards. 

When the first stage of adem's learning - learning to understand and operate with perceptions of people involved in a specific kind of social activity - is completed, adem is tested on a sample of perceptions belonging to people who, although involved in the same kind of activity, did not participate in the first stage of learning. If the results are found satisfactory by the adem's teachers, the second stage of learning is ready to start. 

An example of practical realization of the first stage is any successful accomplishment of a fuzzy inquiry on the Internet. The inquiry represents a fuzzy individual perception to be understood by the world wide web; if the inquirer is satisfied with the search result offered by the server-in-use, it means that the web has adequately understood his/her perception. 
 

top | menu

6. Learning to Create

At the second stage, adem learns how to use its ability to understand human perceptions in order to 'create' autonomous decisions. An autonomous decision represents a meaningful combination of perceptions already understood by adem. 

The meaning of any successfully understood human perception is saved as a specific configuration of adem's neural network. By firing more than one configuration and performing different operations with them, similar to the operations used in the first stage of learning, it is possible to create new meaningful combinations. Whether the new combination is meaningful or not is decided by a teacher. Each newly created meaningful combination is considered as a carrier of a new meaning, and therefore as an approved autonomous decision created by adem. (In a similar way, students proceed while trying to create new meanings in their essays while using previously accumulated knowledge and the assessment of their teachers.) 

This process of learning can be significantly accelerated, if a web of adems is built. Then each autonomous decision generated at one only node of the web and approved as a meaningful can be immediately used at any other node for generation of new autonomous decisions. In this way a kind of a chain reaction of an expanding set of newly created autonomous decisions can be fired; this increases the efficiency of the learning process. 

Example of adem's way of creating autonomous decisions one can find in any Internet-based learning environment, where certain key statements in the teaching material (that is, statements whose meanings are with high priority for understanding the material) are automatically re-shaped into questions and then offered to students as a controlled test for assessment. Students' fuzzy perceptions of the teaching material - perceptions expressed in their responses to the test questions - are processed using techniques applied by adem during the first stage of its learning. Basically, each answer is compared with the corresponding key statement form the teaching material, and if the meaning extracted from the answer coincides with (or is closed to) the meaning of the key statement, the answer is classified as right.

top | menu

7. Conclusion

By focusing on the factors and conditions facilitating the procees of learning, learning ecology opens new possibilities for realization of the self-organizing ability both of the human and the machine intelligence - an ability that plays a crucial role in the creative decision-making. The development of the SC capacity for understanding and dealing with the fuzziness inherent in the holistic nature of the human perceptions is of a major significance for stimulating the emergence of an ontological leap of the SC-based intelligence from its use as a supporter of human decision-making to its use as a 'full-blood' creator of autonomous decisions. This leap is inevitable on the way to the ever-strengthening symbiosis between the human and the SC-based intelligence.

top | menu

References

top