8.3 Lattice Artificial Chemistry

A significantly different approach to computational autopoiesis, termed Lattice Artificial Chemistry (LAC) has been recently pioneered in a series of studies by Ono and Ikegami.

LAC differs from the original model of Varela et al. [46] (and its direct descendants above) in adopting a coarse grained approach. In the original model, each lattice site can generally be occupied by at most one particle: it is fine-grained to the single molecule level. In LAC, by contrast, each single lattice site is permitted to host many particles, of many distinct molecular species.

Using this approach Ono and Ikegami have progressively reported:

Bounded, self-maintaining (autopoietic) cells in one-dimension; reproduction of these cells through two distinct mechanisms [37].
Bounded, self-maintaining (autopoietic) cells in two-dimensions; reproduction of these cells [36,38]. A particular innovation here was the introduction of anisotropic repulsion among one particle species. This is critical to the formation of ``thin films'' which can function as membranes.
Variant strains of two-dimensional cells are introduced, based on a variety of individually autocatalytic particles, with different rates of membrane particle generation [39]. ``Mutation'' between the varieties of autocatalysts is possible. Experiments are reported in which there is selection at the proto-cell level, favouring cells with the higher productivity autocatalysts, and an optimised cell size.
Most recently, a generalization to a three-dimensional space has been described (with Madina, [21]).

Because LAC uses the coarse grained approach, direct comparisons with the fine-grained models are difficult. For example, when multiple cells are present in the same space, they normally share membranes, so that a population of cells exhibits a ``honeycomb'' structure. It is difficult to assess the degree of ``individuation'' on the part of single cells in this case (as multiple cells contribute to the maintenance of any given boundary).

The exhibition of selection at the cell level seems particularly significant [39]. At this point, this is still limited to selection among a pre-designed variety of cell strains--there is nothing like a von Neumann style genetic architecture, or programmable constructor. Nonetheless, the LAC methodology clearly presents a very fruitful avenue of research.

Barry.McMullin@dcu.ie