Bart G

I am a computational social science researcher. I hold a PhD in Industrial Engineering from the University of Toronto.

[bgajdero]

Interview: A Simulator that Predicts the Impact of Social Service Policy

My supervisors and I were recently interviewed about my project, the “A Simulator that Predicts the Impact of Social Service Policy,” at the Centre for Social Services Engineering.

You can read the full inteview here: http://csse.utoronto.ca/interview

Summary

University of Toronto researchers are developing a simulation tool that will predict the impact of social service policies on society's most vulnerable populations. Researchers at U of T’s Centre for Social Services Engineering are exploring innovative technologies designed to improve the way social services are evaluated and delivered to vulnerable populations. Of particular interest are the homeless, at-risk youths, seniors and families living below the poverty line. The Centre is one of the first of its kind to apply engineering techniques and methodologies to the social services sector. One of the Centre’s groundbreaking projects is a computer simulator that can model a city’s network of social services and track specific groups of clients as they make decisions and move through a simulated social support system. The multi-disciplinary research team supporting this project includes U of T and Centre faulty members Dr. Vicky Stergiopoulos, Assoc. Professor & Director, Department of Psychiatry and Professor Marion Bogo, Factor-Inwentash Faculty of Social Work.

I recently spoke with the Centre’s director Dr. Mark Fox (MF), Dr. Michael Gruninger (MG) and PhD candidate Bart Gajderowicz (BG), the engineering research team spearheading the Social Services Simulator project.

[bgajdero]

The passing of Marvin Minsky

Source: http://phys.org/news/2016-01-marvin-minsky-artificial-intelligence-dies.html

On January 24, 2016, Marvin Minsky passed away. Along with John McCarthy (who passed away in 2011), Minsky was the co-founder of the field of Artificial Intelligence. 

I searched for news articles and blog posts to find something to share on social media that would describe what Minsky’s and McCarthy’s work meant for computer scientists like me, and specifically ones working on artificial intelligence, and convey that to someone who does not fall into either of those groups. Listing the numerous contributions Dr. Minsky made to different fields would be great, but iterating through lists is as meaningless now a days as having meaningful conversations with Google in hand, a forgettable experience.

I’d much rather people understood how Dr. Minsky viewed the world, how this shaped the understanding of our own worldview, and what were his motivations for having the views he expressed so eloquently.

What I learned from Marvin Minsky

If I could summarize my views on Marvin Minsky’s life, his goal was not to simply find answers, but to understand the purpose behind the question, and have that lead you to what sub-questions are most interesting. Well intentioned work with no purpose is a tough direction to pursue, and persevere through. In my master’s I looked at two fields within computer science, machine learning and discrete math, to see how together they can find a solution to a the problem of system integration. In my PhD research, I look to engineering and computer science to see how we can help with social problems like homelessness and the psychology of motivation. In my master’s the purpose was to see what I can learn from combing those two fields. In my current work, I’m bridging the very big and very real gap between industrial engineering methods and the problem of homelessness. Taking a step back to see the landscape of problems allowed me to see gaps in knowledge and possible solutions that span multiple areas.

Health, population and the human mind

In 2003, Minsky gave a TED titled Health, population and the human mind (13 minutesvid 1). He gave an extended version of this talk in the second video below. The talks cover two of his books, The Society of Mind (1988) and The Emotion Machine (2006)

13 minute TED Talk:

An extended and more honest 52 minutes talk at the Centers for Disease Control and Prevention (CDC)

A few articles I thought captured who he was and what he meant.

In Honor of Marvin Minsky’sContributionson his 80th Birthday (2007) [PDF]

In this article, published by AAAI in 2007, several scientists who knew him speak about his contributions to AI and their own experiences with him.

“Marvin came to Princeton as a graduate student in mathematics in 1950, a year after me. We quickly realized that we both were interested in artificial intelligence, as we later came tocall it. Unlike me, I think Marvin already had some definite ideas about how to achieve (human-level) AI. These then resulted in his design for the SNARC neural net learner and later led to his 1954 Ph.D. thesis. I had hadideas different from his but didn’t consider them good enough and didn’t come to a definite approach (through logic) until 1957. Neither approach has yet reached human level—nor have any of the others.”

- John McCarthy

The Many Minds of Marvin Minsky (R.I.P.)

An interview with the well intentioned “science sceptic” John Horgman gives us some insights as he asked some thorny questions, and Minsky’s responses and reactions show a man still deeply enthralled by big ideas and deep investigations.

“Minsky also wanted to know what Yo-Yo Ma, the great cellist, feels like when playing a concerto. But to my surprise, Minsky doubted whether such an experience is possible. To feel what Yo-Yo Ma feels as he plays, Minsky explained, he would have to possess all Yo-Yo Ma’s memories. He would have to become Yo-Yo Ma. But in becoming Yo-Yo Ma, Minsky would cease to be Minsky.”

What Marvin Minsky Still Means for AI

“Genetics seemed to be pretty interesting, because nobody knew yet how it worked,” he said. “But I wasn’t sure that it was profound. The problems of physics seemed profound and solvable. It might have been nice to do physics. But the problem of intelligence seemed hopelessly profound. I can’t remember considering anything else worth doing.”

- Marvin Minsky

[bgajdero]

Getting through Gödel, Escher, Bash

Douglas R. Hofstadter’s “Gödel, Escher, Bash: and Eternal Golden Braid,” or GEB, is a monster of a text, at 760 pages, and absolutely worth the read. If you haven’t yet, I suggest you take a look. At its core the book is about developing two ideas:

Mathematics, visual arts, and music represent the same concepts.

Due to their complexity and our different modes of perception, we understand them differently.

The book gives the reader a mechanical way of extracting meaning from the meaningless, and provides tools for manoeuvring and removing the boundaries that separate these two seemingly distinct facets of our environment.

It took me five months to get through it a few years ago, and it’ll take me many more to fully understand and apply its core ideas, but here’s where I’m at now. Through careful examination, meaningless symbols can gain meaning. Any paradoxes we’ve come across along the way to understanding our environment helped developed scientific enquiry and rigour that guide us further.

Hofstadter, as the title implies, connects mathematics, visual arts, and music by mapping concepts in the work of Gödel, Escher, and Bash, and by relying on the thread that runs through each; they are based on formal systems with rules for constructing and interpreting artifacts within that system, resulting in a single system that is the backbone of all three. The rigour and beauty of each individual domain shines through, and makes this book easy to follow, provided you you have the patience to get through it. As I move to formally define the system I’m developing for my PhD, I decided to revisit this book with the help of these MIT lectures.

Lecture 1: https://youtu.be/lWZ2Bz0tS-s

Lecture 2: https://youtu.be/HqmUuHnvJ98

Lecture 3: https://youtu.be/86AHsIduncM

Lecture 4: https://youtu.be/KbTrDBmXX1U

Lecture 5: https://youtu.be/PBDQL7hp7gk

Lecture 6: https://youtu.be/ko6kkJ7Li5Q

Lecture 7: https://youtu.be/EfeBmRR-S64

[bgajdero]

A professor spent 36 hours living on the street and I got 4 years to see how we can help.

When my supervisor, Prof. Fox first told me he spent 36 hours living on the street to understand what we’re up against, I thought I misheard him. We were on our way to meet a Case Worker at a local shelter to discuss their needs. He asked me if I’ve tried their hot lunch, the one they serve twice a day for the thousands that visit each year. As a volunteer and support worker at the shelter I have tried their lunch, but he meant experiencing the full encompassing process: lining up, finding a seat, going back on the street wondering where your next meal will come from. Think of the last time you thought about your next meal right after finishing a hot and nutritious lunch. These kinds of insights sometimes require first hand experience that engineerings and researchers outside of psychiatry and social work don’t often have.

This week I attended a great talk titled “Bringing Technical Knowledge in Public Policy." by Peter Wallace, Toronto’s City Manager, on how engineers can help the city’s 33,000 employees help its 3.9 million citizens. The audience of engineers was overwhelmingly on board.

Today I’m attending the “Beginning the End of Homelessness in Canada” by Tim Richter, the President and CEO of Canadian Alliance to End Homelessness. 

The new Liberal government is committed to poverty reduction.

The CSSE has a mission to take these issues on with an engineering perspective, to understand social services in the same way other organizations have in the past. When industrial engineers look at how hospitals function, the approach is to model as many aspects as required to sufficiently emulate the process and study its inner-workings. The Lean Methodology developed by Toyota in the 60s and 70s was developed to optimize car production, and has been extended to hospital care as well as running an efficient, vibrant, and focused startup. As a PhD student and the director of the Social Service Chain Modelling and Evaluation project, my role is understanding what factors impact a client’s progress through the system, and how to get them to the next step towards existing homelessness.

My point here is this: When we talk about needing to do something about homelessness in our city we’re not just playing lip service to the issues, or describing a scenario in strictly academic terms.

Toronto has both the will and the way. Let’s keep it moving forward!

[bgajdero]

Optimal happiness is a contradiction, so focus on new experiences

This last post is part five of a short series on social psychology, made up of thought papers I wrote for the Affective Science (PSY5430) course at the University of Toronto, Department of Psychology.

In my posts on Basic Emotions and Psychological Construction, I made the point that the brain, just like the body as a whole, is a self regulating system. The previous posts presented this view in a more general sense, with a focus on the physical brain, and its interaction with the neural activity related to emotion processing. This week’s topics, happiness and well-being, struck me as being a specific example of the self regulating brain, but in a semi-subjective way. The duality of happiness is its balance between its hedonic aspects, mainly “more positive” verses “less negative” affects [Gruber et al., 2011]. I stated that this is a semi-subjective view because the third component, life satisfaction, is a cognitive aspect which people have some control over and understanding of (e.g. goal attainment). I think it’s safe to say that the majority of well balanced and reasonable people have happiness as a goal, however unattainable those perceived sources of happiness may be. In fact, research has shown that most people, in general are “somewhat happy” and experience “mild pleasant affect most of the time” [Diener and Lucas, 2000]. As the saying goes, “everything in moderation.”

Optimal happiness is a contradiction

The idea of optimal happiness [Averill and More, 2000] is a somewhat contradictory term, at least in the way we use the word “happiness” for two related, but separate ideas. The first definition is the discrete and conscious feeling of intense happiness. This is what we feel at a particular moment when we do something that makes us feel good (e.g. scoring well on a test, winning a match, etc). The second definition is the continuous process of maintaining well-being, the “mild pleasant affect” state mentioned by Diener et al. This process is a balance between feeling happy and unhappy. It’s not necessarily obvious to us that being unhappy, or at least less negative, would be beneficial. Gruber et al. make the point that to sustain happiness we need to change what makes us happy, especially when that happiness is strictly focused on ourselves.

Cycle of social happiness always leaves someone in need

Source: http://news.163.com/photoview/05RQ0001/18822.html#p=7JEURS7K05RQ0001

We are social beings, and it would make sense that the happiness of others would be a more prolonged source of happiness. As it turns out, in fact, while some focus on themselves, they sacrifice the happiness of others [Averill and More, 2000], so it’s very unlikely that there will ever be no one to make sacrifices for. A pattern emerges where one group’s self-directed discrete happiness creates opportunities for a second group’s continuous well-being, at the cost of a third group who are forced to make sacrifices. So in fact, it seems that perhaps we cannot be truly happy unless there are others that suffer, and that is a sad fact. The only way this system would be “fair” if the members of the three groups get to move between the groups. If there’s a political and economic system which does that, I’d like to here about it.

References

[Averill and More, 2000] Averill, J. R. and More, T. A. (2000). Happiness. In Handbook of emotions, chapter 42.

[Diener and Lucas, 2000] Diener, E. and Lucas, R. (2000). Subjective Emotional Well Being. In Handbook of emotions, chapter 21.

[Gruber et al., 2011] Gruber, J., Mauss, I. B., and Tamir, M. (2011). A Dark Side of Happiness? How, When, and Why Happiness Is Not Always Good. Perspectives on Psychological Science, 6(3):222–233.

[bgajdero]

Emotion Regulation, a Teleological Perspective

This post is part four of a short series on social psychology, made up of thought papers I wrote for the Affective Science (PSY5430) course at the University of Toronto, Department of Psychology.

People are teleological, meaning that they are goal directed. Goals are responsible for deciding what actions to pursue and are the driving force in a clients decision process [Ford, 1995][Chulef et al., 2001]. Hobbs et al. declared that humans are intentional beings, encompassing our necessity to have goals, develop plans to achieve those goals, execute those plans, monitor execution to see if expectation has been reached, and modify plans to execute new plans [Hobbs and Gordon, 2010]. Simple goals do not require extensive planning, and the process from motivation to execution is relatively short. Complex goals require a temporal element as they need to be spread out over some period of time, broken up into multiple sub-goals, and planned out. This is what Tamir describes as our ability to postpone immediate pleasures for future benefits, and characterizes the difference in our motivation as wanting verses liking [Tamir, 2009]. Similarly, Ochsner et. al states that we are able to selectively enable focus on goal oriented stimuli and ignore goal-irrelevant stimuli [Ochsner and Gross, 2005]. My research has identified a number of requirements that need to be satisfied in order to model human behavior [Gajderowicz, 2014]. The teleological elements we identified were goals, expectations, need for sub-goals, goal temporality, and planning.

The underlying brain processes which allow us to accomplish such goals are known as Emotion Regulation (EF) [Gross, 1998]. Its overall purpose is to ensure our brain stays at a stable state, where an unstable state occurs when there is a discrepancy between our goal state and our current state [Zelazo and Cunningham, 2009]. The levels of consciousness grow in parallel with complexity levels required to achieve a goal. Quick and simple decisions do not require planning or the creation of sub-goals, so low levels of consciousness are activated. These goals can be resolved using what Zelazo et. al call cool EF. while complex goals require the hot EF. In this system, complex goal-initiated problems are neutrally contextualized by the hot EF system in a recursive fashion to be processed by the cool EF system.

An abstract representation of this process is a probability tree which represents each permutation of available choices. Each level in the tree is a level of consciousness required to compute different choices. As the decision making process moves down the tree towards the leaf nodes, the goal becomes less abstract. At the leaf nodes, the final decisions are represented and processed by the cool EF system. By seeing more of this tree, the algorithmic level allows the brain to identify problem spots and focus on those [Zelazo and Cunningham, 2009].

This reflects the focus evolution has placed on negative information as a defence mechanism. Whether this is because negative information is more significant or that the brain puts more significants on negative information is still an open question.

References

[Chulef et al., 2001] Chulef, A., Read, S., and Walsh, D. (2001). A hierarchical taxonomy of human goals. Motivation and Emotion, 25(3):191–232.

[Ford, 1995] Ford, M. E. (1995). Motivation and Competence Development in Special and Remedial Education. Intervention in School and Clinic, 31(2):70–83.

[Gajderowicz, 2014] Gajderowicz, B., Fox, M. S., Grüninger, M. (in press). Requirements for an Ontological Foundation for Modelling Social Service Chains. In Y. Guan, J. Liao (Eds.), Proceedings of the 2014 Industrial and Systems Engineering Research Conference. Montréal, QC.

[Gross, 1998] Gross, J. J. (1998). The Emerging Field of Emotion Regulation : An Integrative Review. Review of General Psychology, 2(3):271–299.

[Hobbs and Gordon, 2010] Hobbs, J. R. and Gordon, A. S. (2010). Goals in a Formal Theory of Commonsense Psychology. In Proceedings of the Conference on Formal Ontologies in Information Systems, pages 59–72.

[Ochsner and Gross, 2005] Ochsner, K. N. and Gross, J. J. (2005). The cognitive control of emotion. Trends in cognitive sciences, 9(5):242–9.

[Tamir, 2009] Tamir, M. (2009). What Do People Want to Feel and Why? Current Directions in Psychological Science, 18(2):101–105.

[Zelazo and Cunningham, 2009] Zelazo, P. D. and Cunningham, W. A. (2009). Executive Function: Mechanism Underlying EMotion Regulation. In Gross, J., editor, Handbook of Emotion Regulation, chapter 7. Guilford Publications Inc.

[bgajdero]

Does the Brain Have an Emotion Module?

This post is part three of a short series on social psychology, made up of thought papers I wrote for the Affective Science (PSY5430) course at the University of Toronto, Department of Psychology.

This week’s post focuses on the physiological aspect of emotions, mainly the investigation into the neurological configuration of the brain which deals with affect, appropriately called Affective Neuroscience.

Our original understanding of a living organism’s body, especially primates, is the physical configuration of various body parts, and the obvious functions they perform. The heart is clearly made for pumping blood throughout our system. The human legs are meant to keep us upright and give us the ability to move about. The eye’s most obvious function is to convert light into signals our brain can interpret. This organ-function view of the body has been applied to the brain as well, which is the locationst view [Lindquist et al., 2012]. With the advancement in tools, we discovered that the brain is not simply a set of modules, but a network of neurones. The obvious question, then, is whether the brain could be understood as both individual “parts”, analogous to a leg or an eye, as well as a network of connections. This is the argument of whether a brain is modular, a network, or both, which is the constructionist view [Lindquist et al., 2012].

If we consider the constructionist view of the brain, an internal modularity can still exist in both the physiological and functional forms. Functional in the sense that a group of neurones not necessarily spatially close, work together to perform a task. The formation of modules and hubs has been observed in artificial neural networks when various real-life constraints are placed on the communication between nodes [Pan and Sinha, 2007]. This research has found that modules are formed naturally when a cost is placed on the spatial distance (based on edge-counting) between nodes. This is a realistic view of a physical organ which must balance energy consumption and effectiveness.

Consider the amygdala, a physical hub, connecting various external stimuli to parts of the brain. The amygdala is closely related to fear, meaning fear should be correlated with those senses [Janig, 2003]. Its function has to do with quick responses and interpretations, so it makes sense that the amygdala takes the architectural shape of a tightly connected and efficient hub. In contrast to physical hubs, network hubs in the human brain exist to a serve different type of function [van den Heuvel and Sporns, 2013]. They allow for high levels of functional diversity and functional synchronization between cortical regions. Interpreting emotions is a complicated task and its processing has been observed to be divided between a number of regions with multitude of complex connections. For example, the recognition of emotion from facial expressions and prosody are split between the right and left temporal lobes [Berridge, 2003]. Such hubs form early on in brain development. For example the connection between the medial posterior cingulate, frontal and insular regions are already present in the postnatal infant and child brain [van den Heuvel and Sporns, 2013]. Due to the sheer numbers of neurones in the brain and their complex interconnectedness, it may be impractical, if even possible, to try and identify the functions of individual neurones. Instead, its more feasible to identify sub-networks of neurones which function together to achieve a task. It would also be more practical to identify which hubs of neurones largely function as incoming or outgoing points, the so-called sink or source hubs [van den Heuvel and Sporns, 2013]. It may make sense to call a group of neurones a module, based on their spatial proximity, but it should not be viewed as the only architecture the brain has utilized.

References

[Berridge, 2003] Berridge, K. C. (2003). Comparing the emotional brains of humans and other animals. In Handbook of affective sciences, chapter 3. Oxford University Press, New York, New York, USA.

[Janig, 2003] Janig, W. (2003). The Automatic Nervous System and Its Coordination by the Brain.

[Lindquist et al., 2012] Lindquist, K. a., Wager, T. D., Kober, H., Bliss-Moreau, E., and Barrett, L. F. (2012). The brain basis of emotion: a meta-analytic review. The Behavioral and brain sciences, 35(3):121–43.

[Pan and Sinha, 2007] Pan, R. and Sinha, S. (2007). Modular networks emerge from multiconstraint optimization. Physical Review E, 76(4):045103.

[van den Heuvel and Sporns, 2013] van den Heuvel, M. P. and Sporns, O. (2013). Network hubs in the human brain. Trends in cognitive sciences, 17(12):683–96.

[bgajdero]

Do our emotions have a biological basis in the brain?

This post is part two of a short series on social psychology, made up of thought papers I wrote for the Affective Science (PSY5430) course at the University of Toronto, Department of Psychology.

Source: http://www.npr.org/blogs/health/2013/12/30/258313116/mapping-emotions-on-the-body-love-makes-us-warm-all-over

Our physical body reacts very specifically to many different types of physical stimuli, especially when it comes to pain. Although it may be discreet or continuous, our body has evolved to quickly identify pain, alert our minds and bring attention to the source of the pain, and remove it. Unlike this physical pain-attention system, our minds must interact with an enormous variety of continuous stimuli from the environment, be it visual, auditory, or haptic. To be able to process this complexity of information, our brains are constructed as a network of neurones which are much better suited for this kind of continuous and variable stimuli than a brain which is separated discrete modules. In terms of emotions, Barrett identifies this view of the brain as Psychological Constructivism [Barrett, 2009]. This approach states that psychological states are products that emerge from the interplay of more basic, all-purpose components, and makes a discrete separation between a mechanism (brain) and what it produces (emotions). Once this view was defined, it allowed researchers to separate the physical brain from the objects we perceive, which led to the re-examination of the brain at a lower level with a more empirical view. The interaction between cognition, reactions, and what Russell calls core affects is much more complex than could be processed by individual modules [Russell, 2003]. The evidence to support the idea that emotions as natural kinds is hard to identify, presumably because it simply does not exist [Barrett, 2006].

If a body is a self regulating system, it would make sense for the brain to also self-regulate in order to stay in a state of equilibrium, especially in response to stimuli which disrupt this equilibrium. In a similar way that the brain is notified of a sharp pain in a particular part of the body, the brain can “sense” a change in its equilibrium and react accordingly. Whether the required action is immediate (run from a bear) or prolonged (cope with sad news), there is constant activity to bring the mental state to equilibrium, and its progress constantly re-evaluated. The cyclic nature of this process implies what Cunningham et al. call the Iterative Reprocessing (IR) Model [Cunningham et al., 2013]. Again, in the same way that the intensity of pain is felt with varying degrees indicating urgency, the levels of entropy felt by the IR model indicates urgency with which resources are allocated to achieve equilibrium. The source of non-equilibrium are the differences between expected and actual results. By adopting the constructivist view of non-modular brain functions, and relying on the heterogeneity of the network structure, the IR model is free to adopt an empirically observed interpretation of entropy stabilization.

References

[Barrett, 2006] Barrett, L. F. (2006). Are Emotions Natural Kinds? Perspectives on Psychological Science, 1(1):28–58.

[Barrett, 2009] Barrett, L. F. (2009). The Future of Psychology: Connecting Mind to Brain. Perspectives on psychological science : A Journal of the Association for Psychological Science, 4(4):326–339.

[Cunningham et al., 2013] Cunningham, W. a., Dunfield, K. a., and Stillman, P. E. (2013). Emotional States from Affective Dynamics. Emotion Review, 5(4):344–355.

[Russell, 2003] Russell, J. a. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110(1):145–172.

[bgajdero]

Evolution of Basic Emotions

This post is part one of a short series on social psychology, made up of thought papers I wrote for the Affective Science (PSY5430) course at the University of Toronto, Department of Psychology.

"PicassoGuernica". Via Wikipedia - http://en.wikipedia.org/wiki/File:PicassoGuernica.jpg

The body is a self regulating system, from the way organs function in keeping us alive, to the way the brain consciously and unconsciously regulates communication between those organs. Emotions are a vital part of this system, by helping us regulate internal states including needs for external resources, both physical and psychological. The precise source of emotions within the brain is difficult to identify due to three key reasons. Firstly, elicitors of emotions come from various parts of the brain [Ekman, 1977]. In recent years, much progress has been made to identify regions which elicit neurobiological and neurochemical activity associated with the sensation and recognition of emotions. Secondly, so-called basic emotions are not directly sensed by human beings [Izard, 1992]. What we sense is a combination of “basic emotions”, with varying degrees, or levels of valance [Panksepp, 2000]. The elicitors of those emotions do not trigger them with consistent degrees, if at all. Any elicitor works in combination with other elicitors. Finally, the stimuli eliciting emotions is not consistent. There are many internal and external factors which affect how we interpret stimuli. These factors make it difficult to correctly define the system of emotions, and scientists are looking for strategies to tackle the complexity issue. The categorization of these factors and their analysis within the context of particular categories is a major step in tackling the complexity of our bodies [Panksepp, 2000].

Over time, our needs became more complex, whether through the need to be more innovative in satisfying those needs (harder to acquire resources, etc.) or by the diminishing affect resources have on satisfying those needs (growing family, buildings up tolerance, etc). Adoption to changing environmental conditions is a key evolutionary trait of any organism. For humans, and arguably other mammals, emotions play a key role in setting and successfully acquiring goals through various motivating factors. The need to adopt to a changing environment comes not only through evolutionary changes in our physical bodies, but also in our abilities to learn new skills. Choosing which skills produce maximum outcome does not occur through evolution, although one might argue that the ability to choose does, by developing larger cognitive capacity. There are also social factors which affect our survival rate, such as the ability to care for others and be cared for. This explains why many emotions have an external side that communicates to others our internal states in relation to external goals. Our ability to feel empathy for others, to enact those feelings in other people towards us, and to work together towards a common goal demonstrates the interplay between emotion, motivation, and execution of a plan to achieve an objective. But more importantly it demonstrates the interconnectedness between physical needs, emotional interpretation, and cognitive reaction.

Much confusion in categorizing emotions is caused by apparent overlap of similar emotions which have different elicitors [Ekman, 1977]. For example, some emotions have a defensive function in order to protect ourselves from physical harm. We may have the need for self-preservation from both physical and psychological harm. The triggers are very different but we interpret them in similar ways. Hunger is a physical reaction to our body not having enough resources to function. The uncomfortable feeling of weakness sets in along with physical reactions such as trembling of limbs. This type of physical distress must be perceived emotionally, otherwise we would not have the cognitive reaction to acquire food. The gathering of food is a long term goal with many sub-goals. We must have the cognitive ability to identify that goal, break it up into sub-goals, assess whether we can satisfy those sub-goals, and finally organize them into a comprehensive and realistic plan [Hobbs and Gordon, 2010] [Chulef et al., 2001] [Bargh and Morsella, 2008] [Schank and Abelson, 1977]. This example demonstrates the interconnectedness between physical needs, emotional perception of those needs, and the role cognition plays. Continuing the example of hunger, humans and possibly other primates, developed the ability to persuade others to collect food for us when we ourselves are not able. This developed as our ability to elicit empathy and altruism in others [Panksepp, 2000].

References

[Bargh and Morsella, 2008] Bargh, J. A. and Morsella, E. (2008). The Unconscious Mind. Perspectives on Psychological Science, 3(1):73–79.

[Chulef et al., 2001] Chulef, A., Read, S., and Walsh, D. (2001). A hierarchical taxonomy of human goals. Motivation and Emotion, 25(3):191–232.

[Ekman, 1977] Ekman, P. (1977). From Biological and Cultural COntrobiutions to Body and Facial Movemenet in the Expression of Emotions.

[Hobbs and Gordon, 2010] Hobbs, J. and Gordon, A. (2010). Goals in a Formal Theory of Commonsense Psychology. In FOIS, pages 59–72.

[Izard, 1992] Izard, C. E. (1992). Basic emotions, relations among emotions, and emotion-cognition relations. Psychological review, 99(3):561–5.

[Panksepp, 2000] Panksepp, J. (2000). Emotions as Natural Kinds within the Mammalian Brain. In Biological and Neurophysilogical Approaches, chapter 9, pages 137– 156.

[Schank and Abelson, 1977] Schank, R. C. and Abelson, R. P. (1977). Scripts, plans, goals, and understanding: An inquiry into human knowledge structures. Psychology Press.

[bgajdero]

Frédérick,

Thanks for the great write up. 

I've often used the term folskonomy when introducing people to semantics. Tags are simple concepts to understand, but more specifically, to use. My hope here is that once people start to use folksonomies to categorise content, the leap to using a vocabulary to "tag" more granular ideas in that content will be  easier.

-Bart

[bgajdero]

It's true that for data integration, meshups don't require semantics. This is because integration is performed manually, on concepts that are easily mapped. The most difficult part is the volume of mappings and conversions to perform, not their complexity nor verification. The small number of more-complex definitions are simply done manually as well, albeit with greater detail.

The pervasiveness of semantics is completely overlooked, so much so that they are seen as a black-box without really thinking about the underlining libraries. A simple API call is where a developer's or a project manager's understanding needs to stretch, and no further.

Controlled vocabularies are being used for specific reasons and again API consumers only use the bare minimum of what they require to get the job done. Frédérick Giasson's recent post on ontologies explains the different types of ontologies and how they tie into to the generally accepted understanding. Unfortunately, this understanding usually stops at a folksonomy, a closed-world vocabulary, or Linked Data. Anything above arbitrary semantics is not required for any of these structures.

Your point about the real contribution of Linked Data is spot on. RDF is sufficient for the majority of what people want to use it for. And really, when using it without any semantics, RDF simply becomes a microformat (RDFa) I've discussed the limitations of this elsewhere. RDF can easily represent the relations between relational tables and columns. But it would not suffice to represent the business logic. More expressive languages like OWL are required.

Your point about closed-world vs. open-world is important as well. Because most systems are developed with closed-world assumptions (read defaults), using open world semantics is not required. This is one of the key places where we are in fact asking the wrong questions and looking in the wrong places. When systems step out of their application's domain to perform meshups is the time when they need better semantic technologies. Companies don't often have the need to do this except in incremental steps. The steps ar so incremental in fact, there is no need to invest the resources to make this process smoother. Manual integration and verification is sufficient.

The Ontology Driven Apps post was a great introduction to this methodology brought on by Nicalo Guarino and Michael Uschold. Thanks for that as well!

- Bart

[bgajdero]

The Semantics of Data And Business Logic

The fields of Semantics and Ontologies has consumed my life for the past several years. Why?

They are enormous topics with applicability in every field, not just formal logic or pure academics. They are fundamental to the way we interpret and reason with the world around us, and the amount of information about the world we are to absorb is staggering. Our mobile devices, our activity on the internet, and the increased digitization of our records: have all contributed to the exponential growth in data. Sorting out all this data is an ever increasingly difficult task. And it's not just the growth of data that is increasing but also the way it evolves from it's initial capture-state to how we use it.

Fundamentally, when analyzing data you must determine 3 things:

1) What is the nature of the data? Is it:

user records

patient records

news articles

individual tweets

sensory data from weather stations? 

      All must be treated differently.

2) Are there any patterns in the data?

3) How does the data relate to itself and to other information sources?

What does the data tell us and what can we learn from it? 

Can the correlation of patient symptoms tell us something about the causation of their condition? 

How accurate are weather predictions based on past records? 

Question (1) deals with questions such as (in the weather domain) "What is snow?", "What is temperature?", and maybe even "What is low temperature?". Humans have no problem answering these questions, but machines run into several issues mainly becuase "Snow" is just a 4 character label. This label is not enough to encapsulate all that is "Snow". The technology behind Linked Data begins to reveal this information. At the very least, it provides a point of reference as an URI for an intended meaning of "Snow". This particular DBpedia link provides meta data that is associated with the concept "Snow". Now anyone in any language can reference Snow and mean the same thing. 

Question (2) is addressed by data-mining algorithms, which reveal predictive patterns. For example, the relation between snow and low temperatures is straight forward. The fact that we observe snow becomes a prediction of the current temperature, mainly that it is low. To a computer algorithm, "Snow" is just a label that has 4 characters "S", "n", "o" and "w". A data-mining algorithm might see that whenever the column "condition" is "Snow" the column "temperature" has a value < 0, and derive the rule:

If "condition" == Snow then "temperature < 0"

Question (3) deals with how concepts behind the data relate to each other. Currently this type of knowledge is known in computer systems as "Business Logic". It is the part of software that interprets input and determines what to produce as output. Often this is described as the process between the data and the user interface. Software frameworks that incorporate this distinction are called Model-View-Controller frameworks. Model is the data, Views are the interfaces, and a Controller is the business logic.

Business logic is a set of concrete steps that must occur whenever certain conditions are true. Anyone who is implementing such systems must have very precise specifications, or understand what these concepts represent and how they relate to each other. A system like DBpedia and the appropriate URI will tell you what a concept is, and provides a good amount of metadata. To understand how this concept is related to the rest of the system, it helps to know how it fits into the greater scheme of things.

A) It may exist as a node in a taxonomy, a hierarchical class structure such as:

Snow is a type of Precipitation,

Precipitation is a type of Weather phenomenon,

... and so on.

B) It may describe a process such as:

A "Customer" can withdraw $200 from "Bank Account" if that account's "Balance" >= $200.

C) It may also describe business triggers:

If a website visitor's Ad "X" criteria is matched, display Ad X.

@LambdaVerdonckt has a great post on the main types of ontologies, and their differences.

Once a developer understands what data models will be interacting with a system, the proper functionality may be built around them. It simply is not enough to program a system that "given A, produces B". In today's ever evolving applications, APIs and NoSQL data models, it is not sufficient to develop a system from a set of requirements. The application development cycle must be more dynamic. Agile Software Development is a great way to develop applications in a fast pace, quickly adapting environment. One of the methods of specification gathering is not to receive a set of requirements from a middle man, but to sit face-to-face with the person that is determining the requirements, knows the use cases, and knows who the users are. In other words it's the person who knows the motivation behind the system and what problem the system is suppose to solve. While tying the models to functionality is an in depth topic on its own, having a better understanding of the data beyond columns and relational tables is important in understanding the problem domain.

Interpreting the Data

The two components, data and business logic are currently described by different languages. For example SQL for the data and Ruby for business logic. But is it possible to use a single language for both? Such a language would require enough expressivity to accomplish this. It would require the proper semantics to describe a Client, a Bank Account, and the types of interaction one can have with the other such as Withdrawals, Deposits, Overdrafts, etc.

Ontologies provide this type of functionality. To analyze a data model, we need to recognize individual elements of the model itself.

The field of Mereology deals with elements that are "parts of a whole":

An engine is part of a car.

The field of Set Theory deals with class membership:

This object is a car.... This car is a Honda.

What I've done is express some information about our observation .... "Car", mainly that it has an engine and that it is a Honda. My ability to express this successfully is dependent on the audience understanding what I'm saying. Understanding in this case means two things:

The audience understands what the individual words are.

The audience understands what the individual words mean.

Now that we know some things about a "Car", we can reason about it to infer more information. For example:

The owner of a Car owns the engine inside that car.

We can also infer that: 

The owner of this Car owns a Honda.

See my previous post on the semantics and syntax of these examples.

I've just used a particular syntax to express these relations. Notice that this syntax doesn't say anything about the semantics of these relations, it's just a way to express these statements. To give terms like "Car" meaning, we assign a reference (such as URI) and related concepts (with an ontology). From this we can begin developing the types of interactions a "Driver" can have with a "Car". 

Driver can Drive a Car.

Driver can Steer a Car.

Car has a Break Pedal.

Driver can Press a Break Pedal.

In a broader sense, we call these interactions relations. These types of "common sense" relations can be stored in an upper ontology. Someone developing a "Car" or "Bank" application can use these types of "Driver and Car" relations to better understand the application's domain, and reuse existing relations. If need be, the relation "Drives" can be applied to trucks, and other vehicles. The term "Drives" is related to "Rides" and "Flys", so many of the relations (such as "Steers" can be applied to "Motorcycles" and "Airplanes".

Semantics of Business Logic

Some critics will say that a software package and a database is far more efficient then reasoning with a set of ontological objects and relations. These criticisms are valid, but improvements in semantic technologies have made it possible to adopt semantics into modelling not just data, but also processes embedded in business logic. A recent post by Fiona McNeil titled "The semantic web is here. Is your organization ready?" addresses this type of adoption of semantic technologies.

A reasoner that infers output based on ontology relations and individuals (representing data) can be optimized for a set of queries, called query rewriting. Such concepts have been thoroughly described in the SPARQL community. This is similar to the optimization that occurs during the compilation of a programs source code. The number of relations handled efficiently by any one system has been increasing in the recent years, with a list of triples stores compiled by the W3C. During the compilation process of any programming code, any good compiler will run an optimizer that makes the code more efficient. Consider that Facebook compiles the entire facebook.com application written in PHP into a single compiled binary for runtime efficiency gains.

The dynamic nature of Web 3.0's applications and data requires a more agile set of tools. In the same way that IDE's have made development easier and more efficient, semantics will allow engineers to model data and business logic quicker as well.

[bgajdero]

Semantics - Part 1 - What and Why

This series of posts is all about how semantics can help us deal with information.

The fields of Semantics and Ontologies has consumed my life for the past several years. Why? They are enormous topics with applicability in every field, not just formal logic or pure academics. They are fundamental to the way we interpret and reason with the world around us, and the amount of information we are absorbing about the world is staggering. Our mobile devices; our activity on the internet; the increased digitization of our records: are all contributing factors to the exponential growth in data. Being able to understand this data is becoming an increasingly difficult task. More on data later, for now I present what semantics are in a general sense.

The most general and abstract definition of semantics is "the study of meaning".In this series I will focus on semantics as applied to Logic and focus on its application in Computer Science.

In this context, semantics deal with relations between 2 entities. Take the plus sign "+" in these two examples.

1) 4 + 2 = 6

2) hard work + luck = success

Clearly the semantics of "+" between (1) and (2) are very similar. In (1) we add two numbers to get a number. In (2) we add two nouns to get a noun. How we arrived at the result is different, and depends on the the two elements being added, and what "being added" actually means.

Let's look at some more examples.

3) An engine is part of a car.

Here the semantics of the phrase part-of mean that an engine is somehow attached to the car, or at least is a part which makes up the car.

4) CR-V is a car.

5) CR-V is a Honda.

6) Honda is a car company.

Here the semantics of is-a identify a classification of a CR-V as car, the make of that car as Honda, and Honda as a company. 

The importance of this becomes apparent when trying to infer information from these statements in an automated or systematic way. 

Lets say we want to answer the following questions: 

a) Does the owner of a car own the engine?

b) Does the owner of that car own a Honda?

c) Does the owner of that car own Honda the company?

For humans, it is relatively easy to answer these questions because we can determine the semantics of part-of and is-a from the context of the question. Based on the semantics of part-of we can infer that (a) the owner does indeed own anything that is part of her car. We also infer that (b) she indeed owns a Honda (noun), but (c) she does not own Honda the company (proper noun). Computers don't have this intuition, and must rely on a more formal understanding of these statements.

To talk about how computers can interpret these statements, we need to briefly discuss the difference between semantics and syntax. Syntactically we can view statements 3 - 6 as:

a R b

where a and b are concepts and R is the relation between them. This is called a triple. 

By having the ability to break this down syntactically a machine can ONLY parse these statements into its individual parts. Identifying the three parts of (engine part-of car) is a difficult task in itself, but still does not let a machine answer question (a): whether the owner of the car owns that engine.

This is where semantics plays an important part by giving us some insight about the relation itself. The semantics of part-of tell us that part-of is transitive, meaning that:

(a part-of b) and (b part-of c) then (a part-of c)

This essentially tells us if you own a car, you own every part of that car, including the engine and any part of the engine. We now answered question (a).

This also tells us that if a person owns a car, and that car is-a Honda, that person is also an owner of a Honda. We now answered question (b).

Question (c) is a bit different, and demonstrates the power and complexity of semantics and reasoning in general. Does the owner of a Honda model own Honda the company?  The semantics of the is-a relation change once again. It is still transitive as above, but in a different way. In (b) Honda was a class of a car and a noun, but in (c) Honda is a name of a particular company and a proper noun. What changed is the meaning of Honda, and any relation a concept has to it.

This means that is-a has different transitive semantics for different classes of concepts. In the example above the difference is between a noun and a proper noun. Determining what semantics to apply is a large field of study, which I will attempt to address in future posts. Here is the break down of answering questions (b) and (c) to demonstrate the difference in semantics:

Answering question (b): Suzie owns car X Car X is-a Honda Honda (noun) is-a make of car >> therefore Suzie owns a Honda.

Answering question (c): Suzie owns car X Car X is-a Honda Honda (proper noun) is-a company >> This says nothing about Suzie owning Honda the company.

Analysis: The way we defined Honda changed from a noun to a proper noun, which changed the relation between Suzie and Honda.

In part 2 I will give a more formal definition of semantics, with several resources from academics and industry. I will also provide examples of ontologies with different type of semantics.

[bgajdero]

Launching bartg.org

Welcome to bartg.org, the personal blog of Bart Gajderowicz. Here you’ll find the things that interest me with the occasional rant on everything else. I’d love to hear from you so do contact me via email, by posting a comment or via social links on the right. My previous blog has been archived (see below). 

Who I am …

I am PhD student in the Industrial Engineering Department at the University of Toronto. My supervisors are Mark S. Fox and Michael Grüninger.

I hold a master's degree from the Department of Computer Science at Ryerson University, focusing on semantic integration and machine learning. Up to recently I was the Senior Research Assistant at the Ubiquitous and Pervasive Computing Lab in my department.

What I do …

My interests include an array of topics, but my time is usually consumed by the technical aspects of information/data/text processing, programming, logic, philosophy, and the social engagement and interaction with technology.

My current PhD research focus is modeling the behavior of Social Service clients and practitioners, the process involved in service delivery, and the social context and constraints. you can see my first publication outlining requirements for my work here.

As part of my master's and undergraduate research, I have studied machine learning, and the various ways it is applied to the areas of ontology matching, search, classification, pattern recognition, data mining, pattern analysis, context awareness and new media.

I enjoy philosophy and have spent the last several years consumed by ontologies, semantics, logic, reasoning, inference, and the technical issues that surround their applications. Not a logician by any means, my research interests include semantic technologies, reasoning and machine learning. Recently I have focused on Description Logics, OWL 2, the now many definitions of the Semantic Web, and incorporating these areas into machine learning.

I enjoy programming and although I have currently been using Python in my research, I work mostly with Ruby as my language of choice. Its simple and natural syntax makes it an absolute pleasure to work with, not to mention makes prototyping concepts dead simple and quick. Professionally I have been a Ruby on Rails developer since 2006. I also program in C/C++ and Java, especially when incorporating existing libraries often written in Java. I need to contribute to Open Source projects more often. Me on github.

Analyze things from first principles and base your actions on common sense. For example:

Rhetoric is easy to spot when fundamental principles of argumentation are applied. I have found that in general the problem is usually a difference in semantics and assumptions. I have been called out for rhetoric, and appreciated the effort put into this analyses. Cool heads not only prevail, but also find a way to see each other's reasoning and points.

I don’t program in assembly language, but it does help to understand how a compiler and assembler translate my code into machine instructions; how the efficiency of code affects the stack and the heap; what the 1’s and 0’s mean; why garbage collection is necessary; etc. 

Why am I archiving bartgdev.com?

Because I haven’t blogged anything in over 2 years

My hosting company decided to lose my database and it is currently being served up from cache :(

I like this domain better :)