小蓝视频

Centre for Cognitive Science (COGS)

Spring term 2015

Spring 2015

Tuesdays 16:00 - 17:30

DateSeminarVenue

Jan 27 

Construction Kits for Evolvable Types of Minds 
Aaron Sloman   
University of Birmingham

Jubilee 144 

Feb 3

Semantic and Syntactic Factors Influencing Language Production: "WHO" are we likely to refer to and "HOW" in repeated text?  
Bojana Ivic  
小蓝视频

Jubilee 144 

Feb 10

Small Steps Towards a Computational Model of Insight Problem-Solving
Tom Ormerod
小蓝视频 

Jubilee 144 

Feb 17

The Immorality of Artificial Emotions
Blay Whitby
小蓝视频 

Jubilee 144 

Feb 24

A Theory of Everything in a Mind, Built from Information in Matter and Society 
David Booth 
小蓝视频/Birmingham 

Jubilee 144 

Mar 3

Live Coding, Notation and Improvisation   
Thor Magnusson
小蓝视频

Jubilee 144 

Mar 10

Evidence for Hierarchical Processing of the Watercolor Effect
Ken Knoblauch
Stem-cell and Brain Research Institute

Jubilee 144 

Mar 17

Artificial Care Agents: Insights from Machine Ethics and Machine Consciousness   
Steve Torrance  
小蓝视频

Jubilee 144 

Mar 24

The Brain as a Model of the World
Oron Shagrir 
Hebrew University of Jerusalem

Jubilee 144 

Mar 31

Art and How We See: An artist's perspective
Mark Ware
Brighton & 小蓝视频 Medical School

Jubilee 144 

Apr 7

Easter Break

 

Apr 14

Objective Meaning - CANCELLED
Alan Costall
University of Portsmouth

 

 

Construction Kits for Evolvable Types of Minds 

Aaron Sloman   
University of Birmingham

This is one of the strands of the Turing-inspired Meta-Morphogenesis project, which aims to identify important transitions in types of information-processing in biological evolution and its products, since the earliest life forms, where information was used only for local control and reproduction. Some of the requirements of simple organisms were analysed by Tibor Ganti who showed how chemistry might provide a suitable construction-kit, but later forms of information-processing included increasingly complex and varied uses of information not only about the physical states of organisms, and their immediate and remote physical environments but also about information and information processing activities in the organism itself and in other organisms (e.g. predators, prey, competitors, collaborators, mates, etc.). Some of that information processing requires rapid changes in complex information structures that are best supported by virtual machinery (as happened increasingly in computer-based systems over the last half century). The need to be able to support various kinds of virtual machinery in ever more complex architectures adds to the requirements of the original construction kit. Many detailed requirements are related to "online intelligence", the focus of work on embodied or enactive or situated cognition (the only kind possible for very simple organisms), but more sophisticated requirements have to be met for organisms with "offline intelligence", including abilities to form remote intentions, construct multi-step plans, build explanatory theories, invent and understand stories, and discover the sorts of mathematics that led to Euclid's Elements, over 2.5 thousand years ago. There are reasons to think of evolution as in some ways like a "blind theorem prover", proving theorems about what is possible on the basis of the initial "construction kit".

There are reasons to suspect that currently understood models of computation (e.g. Turing machines, neural nets, evolutionary computation) are not adequate to support some of those kinds of mathematical reasoning (perhaps including discoveries of "toddler theorems" by pre-verbal children). Chemical information processing, already required for the simplest organisms, may add important new features, as Turing seems to have noticed in his last few years.

The talk will present more questions than answers, and constitutes an invitation to join the project.

Expanded online abstract for the talk and project, with additional links and references:

NOTE:

The concept of "information" used here is closer to Jane Austen than Claude Shannon:

Semantic and Syntactic Factors Influencing Language Production: "WHO" are we likely to refer to and "HOW" in repeated text?

Bojana Ivic
小蓝视频

Who are we likely to refer to in the repeated discourse or text? How are we likely to refer to them? This talk presents 5 experiments, looking at which one out of the people that have already been mentioned we are likely to refer to again, and in doing so, are we likely to repeat their full name or use the pronoun. Pronoun resolution is extensively researched topic, and the present research is looking at two previous studies with the conflicting results. One suggests that different factors influence “who” are we likely to refer to and “how” are we likely to refer to them (Fukumura & van Gompel, 2010) and the other that the same factor affects these two questions (Arnold, 2001). In particular, Fukumura and van Gompel, suggest that semantic role that a person mentioned in preceding clause, plays (e.g. stimulus/experiencer), determines who is referred to again, but that the way in which they are referred to (pronoun/name) is determined by the grammatical role (subject or object). Arnold, who looked at the thematic roles of source and goal, suggests that semantic roles influence both “who” we refer to and “how” we refer to them.

In an attempt to understand why these studies produced different results we carried out 5 studies asking people to write continuations to the sentences which introduce two people related by an interpersonal verb. Both implicit causality verbs (following Fukumura & van Gompel) and the source-goal verbs (following Arnold) were used. The 5 experiments were varied in terms of the materials and methodologies used, in experiments 1,4, 5 we used a method of the forced referent, by indicating with an arrow who should participants refer to (thus creating bias-consistent and bias-inconsistent items), and in these experiments the main question was “whether the pronoun or repeated name are used?”. The answer is that the choice of name  vs. pronouns is influenced by the grammatical (subject) or structural factor (position of the antecedent in its clause); it is not clear, however, whether it is the first-mention or the subject of the preceding clause.

In the ‘no arrow’ experiments (exp. 2,3), semantics does seem to influence “who” we are likely to refer to; for the implicit causality items, people tend to refer to the “stimulus” rather than an “experiencer” for both types of verbs (SE/ES). For the Arnold materials, when participants are asked to provide causal continuation (exp. 2) there is no clear effect, but when participants are free to continue the sentence as they wish (cause, consequence, elaboration were the main types of cont.) they choose “goal” over the “source”. This suggests the “goal” preference, rather than straightforward causality bias. When people are free to choose the continuation, for the ES/SE verbs their preferred choice of ending is “cause”, whilst for the GS/SG verbs, causes are less popular as the continuation, and this might allow people to refer to “goals” rather than “sources”.  Further studies will investigate these questions by asking  people to write consequential continuations. Insofar, it seems that slight changes in either items or instructions, affect changes on more levels than one. This might be an approach analogous to Marr’s theory of vision, which suggests that statistical properties of language affect language comprehension and language production at the very basic level.

Small Steps Towards a Computational Model of Insight Problem-Solving

Tom Ormerod
小蓝视频

Recent theories of insight problem-solving are of two kinds: knowledge-based accounts such as Representational Change Theory (RCT: Knoblich et al., 1999), in which problem difficulty is mediated by inappropriate knowledge; and strategic accounts such as Criterion for Satisfactory Progress theory (CSP: MacGregor, Ormerod, & Chronicle, 2001), in which problem difficulty is mediated by search for moves that maximize progress towards a hypothesized goal. Most researchers agree that both knowledge and strategy are essential for capturing insight phenomena (e.g., Kershaw & Ohlsson, 2004), and integrated frameworks have been proposed (e.g., Hélie & Sun, 2010). However, progress is hampered by the lack of executable models of either knowledge-based or strategic mechanisms. In this presentation, some empirical data and a fledgling computational model are presented that show how knowledge and strategy may interact to give rise to common insight phenomena.

The Immorality of Artificial Emotions

Blay Whitby
小蓝视频

Research into the simulation of human-like emotions is a major research theme in robotics, artificial intelligence, and cognitive science. Unfortunately there is a very high probability that the products of this research will be used – and indeed already are sometimes being used – in ways that are clearly unethical.

Artificial emotions research facilitates the development of a technology that can be used to manipulate, exploit, and abuse humans. It is also an area in which there currently exist almost no legal or ethical restrictions. On balance therefore, it is time to scrutinize this area of technology on moral grounds and to abandon the assumption that it is likely to be benefit to humanity.

A Theory of Everything in a Mind, Built from Information in Matter and Society

David Booth 
小蓝视频/Birmingham

Here is the simplest known theory of the conscious and unconscious processes involved in a person’s achievements such as intending, perceiving, reacting and thinking.

(I) The basic principle is that an individual system has mental states which possess quantitative causal powers and are susceptible to influences from other mental states. Mental performance discriminates the present level of a situational feature from its level in an individually acquired, multiple-featured norm (exemplar, template, standard). The effect on output of a moderate disparity between input and norm is scaled in a universal unit of discriminative sensitivity, with the norm’s level being zero.

(II) When one process converts separate sources of input into an output, their discriminative distances from norm are summated. Distinct processes converging on an output combine their discriminations from norm orthogonally. An output may be influenced by other outputs as well as by inputs. Descriptive performance is the influence of one category of input on a verbal output. Reasoning is minimally the effect of one verbal process on another. In deeper mental processing, the influence on a response comes from a concept modulating a description: this process gives the meaning to an emotion or a motive. Descriptive modulation of stimulation corresponds to a bodily sensation or other conceptualized percept. When an output is explained solely by sources of input, that response to the stimulation may be mediated unconsciously.

(III) Development of a human person or an engineered intelligence within physical and communal environments embodies such mental causation within material causation and acculturates that individual’s mind to social causation, by continual updating of memory (learning).

Live Coding, Notation and Improvisation

Thor Magnusson  
小蓝视频

Live coding is a method of writing software where the program is created and altered in real-time whilst running. Live coding can be applied in diverse contexts, but this presentation focuses on music. Emphasising improvisation, live coding in music is a practice merging the act of musical composition and performance into a public act of projected writing. 
In this talk, Thor Magnusson will give a quick introduction to live coding, but then move on to the context of code as musical notation, the problem of the score, and how/why one would perform writing code. Thor will introduce the improvisational features of his Threnoscope live coding system, which implements a micro-language for drone-based microtonal and spatial composition. The system is designed with the aim of removing representation of linear time from music by emphasising stasis or circularity. Through visualisation of harmonic relationships, tuning systems, and multichannel audio, the system affords specific compositional focus that encourages the performer to work with spatial sound and descriptions of harmonic spectra as opposed to scheduling events in time.
Some musical performance might occur.
Thor Magnusson's background in philosophy and electronic music informs prolific work in performance, research and teaching. His work focuses on the impact digital technologies have on musical creativity and practice, explored through software development, composition and performance. He is the co-founder of ixi audio (), and has developed audio software, systems of generative music composition, written computer music tutorials and created two musical live coding environments. Thor Magnusson has a DPhil from Informatics at 小蓝视频, and has recently joined the Music faculty at 小蓝视频.

Evidence for Hierarchical Processing of the Watercolor Effect

Ken Knoblauch 
Stem-cell and Brain Research Institute

Neurophysiological and neuroanatomical evidence supports the idea that information is processed hierarchically in the visual system, and recent theories of neural processing, such as predictive coding, depend on signals propagated up and down neural hierarchies as a substrate of perception.  What kind of evidence can be demonstrated for hierarchical processing in perception?  I consider this question in the context of the Watercolor Effect, a long-range color filling-in phenomenon induced by wavy, chromatic contours.  Based on experiments in psychophysics and functional cerebral imaging, I will summarize evidence supporting a hierarchical model of this phenomenon.

Artificial Care Agents: Insights from Machine Ethics and Machine Consciousness

Steve Torrance
小蓝视频

This talk looks at two “fringe” AI fields that have recently emerged into prominence: Machine Ethics (ME) and Machine Consciousness (MC). We examine them in the context of an important new application area – artificial care agents (ACAs).
A growing number of applications for such agents are foreseen in the fields of medical and social care. As ACAs acquire increasing operational autonomy, there is growing concern that their designs ensure they are “ethically safe”. This is especially so given that many people whose needs they will be designed to serve are among the most vulnerable in society.
This talk will look at how both ME and MC are implicated in this, and on some interrelations between the two research areas.
(This presentation is based on the following: S. Torrance, R. Chrisley, ‘Modelling consciousness-dependent expertise in machine medical moral agents.’ In: S. Van Rysewyk and M. Pontier (eds) Machine medical ethics. Springer, 2014.)

The Brain as a Model of the World

Oron Shagrir
Hebrew University of Jerusalem

I argue that an underlying assumption in computational approaches in cognitive and brain sciences is that the brain is a model of the world in the sense that it mirrors certain relations in the surrounding environment. I will give here three examples. One is from David Marr's computational-level theory of edge-detection. The second one is the computational work on the oculomotor system. And the third one is a Bayesian model of causal reasoning. One might wonder why this brain-as-a-model-of-the-world assumption is so prevalent in computational cognitive science and neuroscience. My proposed answer (for which I will not argue here) is that in these fields computation just means a dynamical process that models another domain. Thus saying that the brain computes just means that its processes models certain relations in another domain, often the surrounding world.

Art and How We See: An artist's perspective

Mark will discuss his artistic practice before and after his stroke and how his stroke has given him new insights into the perceptual process. He will touch upon how he believes the history of art was affected by the invention of photography, leading to thoughts on optical illusions and how we see.
Amongst these ideas he will refer to Madam Tussaud and the death mask of Robespierre, anaglyph 3D photography, Medieval carving, Muybridge and Duchamp, a train arriving at a station and a tennis ball.
The presentation will conclude with an introduction to The Wavelength Project, an Arts Council England funded art/science collaboration between Mark Ware and Professor Hugo Critchley and his colleagues at the Sackler Centre for Consciousness Science.  The Wavelength Project will investigate the effects of natural versus artificial light and sound on the brain.  Professor Critchley will co-present this conclusion to the talk.
Mark Ware is a multimedia artist, based in Devon.  He is both a commercial artist and fine art practitioner.  Mark will offer a presentation of what appear to be disparate topics and ideas that are, on closer inspection, strongly linked by his experience of having a severe stroke 18 years ago at the age of 39.
'Throughout my adult life I have worked professionally as an artist.  My stroke had a sudden and dramatic affect upon my art and the way I navigate and negotiate and think about my life and the environment around me.' 

Objective Meaning

Alan Costall 
University of Portsmouth

Psychologists, when they have not simply ignored objects and places, have largely taken ‘their’ meaning for granted. For example, we are certainly impressed when a child pretends that a pen is a rocket, but we simply take for granted that even a very young child knows what this thing really is, namely, a pen. Such meaning is impersonal, and, in this sense, objective.

However, such ‘canonical affordances,’ although embodied in particular objects, are also socially constituted. They are not simply there. In this session, I want to explore the ways in which canonical affordances have created a ‘world’ of objective meaning, how this might provide a crucial basis for the development and sustaining of communication, and how such stabilized, objectivized meanings somehow survive the remarkable degree of ‘play’ they also afford.

 

COGS loco