The challenge to simulate the brain

October 7, 2015

Michele FariscoIs it possible to create a computer simulation of the human brain? Perhaps, perhaps not. But right now, a group of scientists is trying. But it is not only finding enough computer power that makes it difficult: there are some very real philosophical challenges too.

Computer simulation of the brain is one of the most ambitious goals of the European Human Brain Project. As a philosopher, I am part of a group that looks at the philosophical and ethical issues, such as: What is the impact of neuroscience on social practice, particularly on clinical practice? What are the conceptual underpinnings of neuroscientific investigation and its impact on traditional ideas, like the human subject, free will, and moral agency? If you follow the Ethics Blog, you might have heard of our work before (“Conversations with seemingly unconscious patients”; “Where is consciousness?”).

One of the questions we ask ourselves is: What is a simulation in general and what is a brain simulation in particular? Roughly, the idea is to create an object that resembles the functional and (if possible also) the structural characteristics of the brain in order to improve our understanding and ability to predict its future development. Simulating the brain could be defined as an attempt to develop a mathematical model of the cerebral functional architecture and to load it onto a computer in order to artificially reproduce its functioning. But why should we reproduce brain functioning?

I can see three reasons: describing, explaining and predicting cerebral activities. The implications are huge. In clinical practice with neurological and psychiatric patients, simulating the damaged brain could help us understand it better and predict its future developments, and also refine current diagnostic and prognostic criteria.

Great promises, but also great challenges ahead of us! But let me now turn to challenges that I believe can be envisaged from a philosophical and conceptual perspective.

A model is in some respects simplified and arbitrary: the selection of parameters to include depends on the goals of the model to be built. This is particularly challenging when the object being simulated is characterized by a high degree of complexity.

The main method used for building models of the brain is “reverse engineering.” This is a method that includes two main steps: dissecting a functional system at the physical level into component parts or subsystems; and then reconstructing the system virtually. Yet the brain hardly seems decomposable into independent modules with linear interactions. The brain rather appears as a nonlinear complex integrated system and the relationship between the brain’s components is non-linear. That means that their relationship cannot be described as a direct proportionality and their relative change is not related to a constant multiplier. To complicate things further, the brain is not completely definable by algorithmic methods. This means that it can show unpredicted behavior. And then to make it even more complex: The relationship between the brain’s subcomponents affects the behavior of the subcomponents.

The brain is a holistic system and despite being deterministic it is still not totally predictable. Simulating it is hardly conceivable. But even if it should be possible, I am afraid that a new “artificial” brain will have limited practical utility: for instance, the prospective general simulation of the brain risks to lose the specific characteristics of the particular brain under treatment.

Furthermore, it is impossible to simulate “the brain” simply because such an entity doesn’t exist. We have billions of different brains in the world. They are not completely similar, even if they are comparable. Abstracting from such diversity is the major limitation of brain simulation. Perhaps it would be possible to overcome this limitation by using a “general” brain simulation as a template to simulate “particular” brains. But maybe this would be even harder to conceive and realize.

Brain simulation is indeed one of the most promising contemporary scientific enterprises, but it needs a specific conceptual investigation in order to clarify its inspiring philosophy and avoid misinterpretations and disproportional expectations. Even, but not only, by lay people.

If you want to know more, I recommend having a look at a report of our publications so far.

Michele Farisco

We like challenging questions - the ethics blog


Where is consciousness?

May 26, 2015

 

Michele FariscoWould it be possible to use brain imaging techniques to detect consciousness and then “read” directly in people’s brains what they want or do not want? Could one, for example, ask a severely brain injured patient for consent to some treatment, and then obtain an answer through a brain scan?

Together with the philosopher Kathinka Evers and the neuroscientist Steven Laureys, I recently investigated ethical and clinical issues arising from this prospective “cerebral communication.”

Our brains are so astonishingly complex! The challenge is how to handle this complexity. To do that we need to develop our conceptual apparatus and create what we would like to call a “fundamental” neuroethics. Sound research needs solid theory, and in line with this I would like to comment upon the conceptual underpinnings of this ongoing endeavor of developing a “fundamental” neuroethics.

The assumption that visualizing activity in a certain brain area can mean reading the conscious intention of the scanned subject presupposes that consciousness can be identified with particular brain areas. While both science and philosophy widely accept that consciousness is a feature of the brain, recent developments in neuroscience problematize relating consciousness to specific areas of the brain.

Tricky logical puzzles arise here. The so called “mereological fallacy” is the error of attributing properties of the whole (the living human person) to its parts (the brain). In our case a special kind of mereological fallacy risks to be embraced: attributing features of the whole (the brain) to its parts (those visualized as more active in the scan). Consciousness is a feature of the whole brain: the sole fact that a particular area is more active than others does not imply conscious activity.

The reverse inference is another nice logical pitfall: the fact that a study reveals that a particular cerebral area, say A, is more active during a specific task, say T, does not imply that A always results in T, nor that T always presupposes A.

In short, we should avoid the conceptual temptation to view consciousness according to the so called “homunculus theory”: like an entity placed in a particular cerebral area. This is unlikely: consciousness does not reside in specific brain regions, but is rather equivalent to the activity of the brain as a whole.

But where is consciousness? To put it roughly, it is nowhere and everywhere in the brain. Consciousness is a feature of the brain and the brain is more than the sum of its parts: it is an open system, where external factors can influence its structure and function, which in turn affects our consciousness. Brain and consciousness are continually changing in deep relationships with the external environment.

We address these issues in more detail in a forthcoming book that I and Kathinka Evers are editing, involving leading researchers both in neuroscience and in philosophy:

Michele Farisco

We want solid foundations - the Ethics Blog

 


How can the brain be computer simulated?

October 29, 2014

PÄR SEGERDAHL Associate Professor of Philosophy and editor of The Ethics BlogA computer simulated human brain – that undoubtedly sounds like science fiction. But the EU flagship project, the Human Brain Project, actually has computer simulation of the brain as an objective.

What will be accomplished during the ten years that the project is financed will presumably be simulations of more limited brain functions (often in the mouse brain). But the proud objective to simulate the human brain has now been formulated in a serious research project.

But what does “computer simulation of the brain” mean?

In an article in the journal Neuron Kathinka Evers and Yadin Dudai discuss the meaning of simulation of the brain. Kathinka Evers from CRB leads the philosophical research in the EU Project and Yadin Dudai is a neuroscientist from the Weizmann Institute of Science who also works in the project.

The article combines philosophical and scientific vantage points to clarify the type of simulation that is relevant in neuroscience and what goals it may have. Several of the questions in the article are relevant also for the simulation of more limited brain functions. For example, the question if the ability to make a computer simulation of a brain function means that you understand it.

The most thought-provoking questions, however, concern the big (but distant) goal to simulate a whole human brain. Is it possible in principle, given that the brain is embedded in the body and is in constant interaction with it? Is it possible, given that the brain interacts not only with the body but also with a social environment?

Does simulating the brain require that one also simulates the brain’s interaction with the body and the social context in which it operates? Kathinka Evers thinks so. The attempt to simulate the brain is too limited if one does not start out from the fact that the brain is in constant interaction with an environment that constantly changes it.

The brain must be understood (and simulated) as an “experienced brain.”

Suppose that one day one manages to simulate an experienced human brain in intensive interaction with a bodily and social environment. Has one then simulated a brain so well that one created consciousness?

The questions in the article are many and breathtaking – read it!

Pär Segerdahl

We like challenging questions - the ethics blog


%d bloggers like this: