Organizational Consciousness Versus Artificial Consciousness

Abstract

What is the capacity of an informal network of organizations to produce answers in response to complex tasks requiring the integration of masses of information designed as a high-level cognitive and collective activity? Are some network configurations more favourable than others to accomplish these tasks? We present a method to make these assessments, inspired by the Information Integration Theory issued from the modelling of consciousness. First we evaluate the informational network created by the sharing of information between organizations for the realization of a given task. Then we assess the natural network ability to integrate information, a capacity determined by the partition of its members whose information links are less efficient. We illustrate the method by the analysis of various functional integrations of Southeast Asian organizations, creating a spontaneous network participating in the study and management of interactions between health and environment. Several guidelines are then proposed to continue the development of this fruitful analogy between artificial and organizational consciousness (refraining ourselves from assuming that one or the other exists).

Annex

The corpus C _S associated with a network S composed of M organizations is built as the concatenation of the individual corpora C _m associated with each organization, symbolically:

$$ {C}_S={\bigcup}_{m=1}^M{C}_m $$

(3.1)

The posterior ¹⁸ probability of occurrence of the term ${\boldsymbol{t}}_{\boldsymbol{n}}^{\boldsymbol{j}}$ associated with organization X _j in the corpus C _m is given by

$$ P\left[{t}_n^j\left({C}_m\right)\right]={F}_{oc}\left[{t}_n^j\left({C}_m\right)\right]/{F}_{oc}\left[{t}_n^j\left({C}_S\right)\right] $$

(3.2)

where F _oc[x] is the number of occurrences of event x. The posterior probability of joint occurrence of two terms, ${t}_n^j$ of X _j and ${t}_{n\prime}^l$ of X _l, in the corpus C _m is:

$$ P\left[{t}_n^j\left({C}_m\right),{t}_{n\prime}^l\left({C}_m\right)\right]={F}_{oc}\left[{t}_n^j\left({C}_m\right)\wedge {t}_{n\prime}^l\left({C}_m\right)\right]/{F}_{oc}\left[{t}_n^j\left({C}_S\right)\wedge {t}_{n\prime}^l\left({C}_S\right)\right] $$

(3.3)

The average mutual information between organizations X_j and X_l in the network S is estimated as:

$$ {I}_{AMI}{\left[{X}_j,{X}_l\right]}_{C_S}={\left({N}_j\times {N}_l\right)}^{-1}{\sum}_{m=1}^{M+}{\sum}_{n=1}^{N_j}{\sum}_{n^{\prime }=1}^{N_l}{e}_{n{n}^{\prime}}^{jl}(m) $$

(3.4)

N _j (resp. N _l) being the number of terms in X _j (resp. X _l). If we consider an additional corpus C ⁺ then M+ = M + 1 (otherwise M+ = M). The elementary information ${e}_{n{n}^{\prime}}^{jl}(m)$ between terms ${t}_n^j$ and ${t}_{n\prime}^l$ on corpus C _m is given by:

$$ {e}_{n{n}^{\prime}}^{jl}(m)=P\left[{t}_n^j\left({C}_m\right),{t}_{n\prime}^l\left({C}_m\right)\right]\ln \left\{\frac{P\left[{t}_n^j\left({C}_m\right),{t}_{n\prime}^l\left({C}_m\right)\right]}{P\left[{t}_n^j\left({C}_m\right)\right]P\left[{t}_{n\prime}^l\left({C}_m\right)\right]}\right\} $$

(3.5)

The elementary information ${e}_{n{n}^{\prime}}^{jl}(m)$ is not zero (and therefore${I}_{AMI}{\left[{X}_j,{X}_l\right]}_{C_S}\ne 0$) if term ${t}_n^j$ of X _j and term ${t}_{n\prime}^l$ of X _l are both occurring in the same corpus C _m (whatever the value of m). The auto information ${I}_{AMI}{\left[{X}_j,{X}_j\right]}_{C_s}$ is not zero (self-loop on the graph) if at least one key term of X _j appears at least in one other corpus C _l, with l ≠ j. The average mutual information between components S _κ and S _π of the S network bipartition is

$$ {I}_{AMI}{\left[{S}_{\kappa },{S}_{\pi}\right]}_{C_s}={N_{\kappa \pi}}^{-1}{\sum}_{m=1}^{M+}{\sum}_{n=1}^{N_{\kappa }}{\sum}_{n^{\prime }=1}^{N_{\pi }}{e}_{n{n}^{\prime}}^{\kappa \pi}(m) $$

(3.6)

the elementary information ${e}_{n{n}^{\prime}}^{\kappa \pi}(m)$ being given by an equation similar to (3.5) except that term ${t}_n^{\kappa }$ (resp. ${t}_{n\prime}^{\pi }$) is taken in the list (composed of N _κπ term pairs$\left[{t}_n^{\kappa },{t}_{n^{\prime}}^{\pi}\right]$) of formed by all key terms of organizations belonging to the bipartition component S _κ (resp. S _π), no term appearing twice in the list. The capacity Φ _IIS[S] of network S to function as an Information Integrative System (IIS) is the amount of effective information that can be integrated across the informational weakest link of a subset of organizations. Limiting the search for this weakest link between bipartition components S _κ and S _π, we have:

$$ {\Phi}_{IIS}\left[S\right]={\mathit{\min}}_{\left[{S}_{\kappa },{S}_{\pi}\right]\in {\Pi}_2(S)}\left\{{I}_{AMI}{\left[{S}_{\kappa },{S}_{\pi}\right]}_{C_S}\right\} $$

(3.7)

Π₂(S) being the set of all bipartitions of S.