Collective intelligence: an interview with Pierre Levy
In a previous post I discussed and, hopefully, debunked some common assumptions on the next phase of the World Wide Web, or web 3.0. The general assumption is that in the 2.0 era the user was at the centre, the produser took control and the cult of the amateur was born. The web was being flooded with what seems an infinite amount of user generated content. Big platforms, such as Flickr and Facebook managed to centralize and collect some of these efforts effectively. The result of is a big, fragmented and messy dataset. Enter web 3.0; the iteration of the web which can be read and understood by machines, where the dots will be connected and contribute to an open sphere of knowledge, something that the current pragmatics of the web don’t easily allow for. The philosophy here is, bluntly put, that this connected sphere is more than the sum of its parts. Tim Berners Lee recognized the problematics of the messy web early on and proposed the Semantic Web to overcome messiness and apply a semantic structure to bring order (read: computer logic) in the chaos (read: human expression).
Pierre Levy, French philosopher and leading expert on collective intelligence, is on a similar mission. While driven by, arguably, a similar set of goals, his approach takes a step further. The problem with Berners Lee semantic structure is that implies a universal ontology, which might prove out to be the Achilles heel of the protocol. Levy’s approach overcomes these problems.
Levy is currently working on a research program, called IEML (Information Economy Meta Language). IEML is a metalanguage and proposes itself as the language of collective intelligence. As a metalanguage it differs fundamentally from natural languages we know. This can be best understood in the way it is conceived. Natural languages are, in the first place, the results of a process of documenting the spoken word. A metalanguage is artificial and is a result of formalizing ideas, instead of words. The practice of formalizing ideas in a universally adopted metalanguage is well established in the realm of natural sciences. For centuries now, ideas are being documented in terms of formulas, numbers, equations, molecules etc. There is a finite, well structured toolset at the hands of every natural scientist. In humanities the area of interest is infinite and not easily encapsulated in a formal manner. In humanities, knowledge is fuzzy, or as the IEML vision paper describes: “the knowledge and expertise accumulated by the humanities are difficult to share in contexts that differ from the initial environment in which they emerged” IEML offers the solution to this problem, offering humanities a language in which knowledge can be formally described. How this exactly will work pragmatically is to be determined, as the project is in a “fundamental research stage”, as Levy stressed out to me. On a hopeful note though: IEML also functions as a bridge between languages, the natural language of the end user is not relevant: “The IEML inter linguistic dictionary is precisely constructed to ensure that semantic networks can be automatically translated from one natural language to another”. Mr Levy was kind enough to answer a number of questions and concerns I had with this project.
As Levy pointed out to me in order for an idea to contribute to the sphere of collective intelligence it should be described in a formal manner. This is where the digital humanities researcher comes in, who should master this new code to formalize his peers ideas, a crucial step in the process: “If ideas and concepts are not formalized, it is impossible to compute their semantic relationships automatically”.
One of my concerns was with the IE in IEML, meaning Information Economy. In the vision paper it is noted that this sphere should be “observed”. I was curious what should be exactly observed in terms of meaningful data and the private sphere, Levy answered:
There is currently an immense mass of public data on the World Wide Web that is not efficiently shared, analysed and used by humanities and social sciences. Considering the extraordinary range of these data and the computing power that is now at our disposal, a scientific revolution in the human sciences can be predicted for the 21st century. I can mention the areas of cultural heritage, health, education, economy, sociology, etc.
One of the main reasons why this computational potential is not actualized today is the lack of semantic interoperability. A universal (interlinguistic and interdisciplinary) system of computable metadata, like IEML, could be the stepping stone leading us into a renewal of human sciences. Of course, every team or individual should be free to categorize and assess the data as he wishes. The common semantic code will allow for comparison and sharing.
All this should be done while respecting existing laws and privacy of individuals. (There is enough work to be done on public data.) The dangers that you mention are not specifically linked to IEML and do exist for all digital data in general.
The scientific revolution in the human sciences will culminate in collective intelligence, a common good which will throttle human development. In a recent book, Levy proposed a “loose IEML model” to monitor the coordination of human development. The axis of human development are defined by “education, health, sustainable economic prosperity, security, human rights, conservation and enrichment of cultural heritage, environmental balance, scientific and technical innovation” which are in accordance with the United Nations Development Program, Levy assures me. I wondered whether a metalanguage which positions itself functionally as neutral (as opposed to Berners Lee universal ontology) should contain assumptions on how western democratic society is structured to which Levy partly agrees that any metalanguage can’t be neutral:
There can be a lot of disagreements about the right ways or methods to improve human development. IEML, as a universal semantic code, can accomodate any method. Above all, IEML provides a common semantic sphere where all disciplines of human sciences can compare their theories and methods and can coordinate their findings at the service of human development. (…) Now, you can say: “Okay, but what if I am against improving health and education because these are western values and / or it has been used to justify western imperialism”. My response is: “It’s up to you!” In general, I do not think that any theory or metalangage can be neutral. Every act, being practical or theoretical, occurs in a hypercomplex context and has an effect on this context. I do not claim any impossible neutrality or objectivity. The objection “you’re not neutral” is besides the point. I have a very precise goal. My aim is to improve human development, collective intelligence and knowledge management in the humanities.
The creation of IEML is based on the explicit assumption that all human beings, and all cultures, have in common a basic linguistic-symbolic ability. The main limitation of artificial intelligence is the belief that logic and statistics are sufficent to model human intelligence. I don’t think that current techniques of automatic reasonning are enough to model the basic symbolic manipulation ability of the human species. In addition to the formal tools of artificial intelligence, we need a new kind of formalism to describe in a functional and computable manner our capacity to create and transform meaning (sense, signification). IEML provides precisely such a formalism. The main result of this scientific invention will be the expansion of a semantic sphere where any creative conversation on line will be able to observe its own processes of collective intelligence and to share methods and results with other conversations. IEML’s existing dictionary will be expanded. You’ll be able to build any kind of “universe of discourse” or semantic world by using IEML. Far from being “neutral”, the creation of IEML points toward a cultural leap forward: a perspectivist scientific reflexivity of human collective intelligence.
My final concern had to do with cultural determination, as Levy had stressed in another interview; the attempts at creating a symbolic metalanguage can be found in many different cultures. Each of these cultures would produce a different “universal” language. How does one overcome this, can a symbolic metalanguage be universal?
What is not culturally determined in the human realm, specially when it comes to language? IEML is first culturally conditionned by the technical environment of the 21st century : growing computing power (automation of symbol manipulation), growing memory power (availability of digital data), and growing communication power (ubiquity of data). It is also conditionned by the scientific method (which is of course a dated cultural institution), namely its insistence on functional computability and transparent (reproductible) procedures. I don’t see these determinations as limitations, but rather as very powerful engines that I have used in the service of human self-knowledge.
Levi-Strauss being one of my favorite author, I am of course well aware of the dangers of “ethnocentrism” . As an inventor, I have been influenced by a wide range of disciplines and theories in the human and cognitive sciences. (All this is explained in my book The Semantic Sphere). I should also mention that I have dwelt on three continents (Africa, Europe, North Amercia) and learned a lot from different cultures, that I have studied traditional Greek, Indian and Chinese philosophy, that I have scrutinized Jewish, Christian, Muslim and Bouddhist metaphysics and that I am deeply involved into the amazing Brazilian cultural and economic development.
Universal does neither mean “out of history” nor “out of culture”. The notation position system of numbers, including the zero, is universal. The decimal system is (almost) universal. The time zones system is universal. The meridian and parallel systems for geography are universal. The Internet Protocol and the HyperText Transfer Protocol are universal. However, all these symbolic systems have been invented somewhere, sometime.”
Finally I enquired about the future of IEML, what can we expect from this project in the foreseeable future. First his team has to build technical tools and train “semantic engineers” the language. After that? Levy is not sure: “Beyond this, I have no precise idea of the actual development [will be]. I just feel that it will happen sooner or later. I think that some big university or some scientific endeavour will lead the way, followed by the companies operating in cloud computing and big data management. I foresee the development of ‘collective intelligence games’ looking like massive multiplayer online / real life games, or some sort of trans-platform smart social media.”
Apart from the technicalities: Levy has a clear vision for the future. IEML can’t be regarded as merely a language or a tool, that wouldn’t do its intentions justice. IEML is a language with a mission and I can’t wait to find out how that will play out.