The Psychophysical World of the Motile Diatom Bacillaria paradoxa

The Psychophysical World of the Motile Diatom Bacillaria paradoxa

Introduction

In the study of organismal behavior from across the tree of life, there are traditionally two types of explanatory mechanism. The first is behavior that is generated by neuronal processes (such as a network of neurons), and is restricted to organisms with a nervous system. The second is behavior that is characterized by loosely-relevant physical, mechanical, and social metaphors and generated by biophysical processes. Yet recently, a new explanatory framework has emerged: non-neural cognition [1-3]. This framework ascribes information processing and decision-making capabilities to systems that lack a formal nervous system structure. Rather, the information processing function occurs in the biophysical substrate itself.

While the diatom Bacillaria does not have a nervous system, it does exhibit organized behaviors and collective movement. Bacillaria consists of a series of filaments arrayed in parallel, and the behaviors exhibited across these filaments are temporally linked in a way that results in coordinated movement. Therefore, we propose that rather than using the multitude of existing metaphors to interpret Bacillaria movement, a model of cognition can be used instead. This model of cognition results in the production of behaviors related to taxis [4] and even psychophysics [5], which can be distinguished from random behavior using a series of criteria.

Four basic regularities of psychophysics as discussed in [6]:

1. regression and range effects leading to the overestimation of weak stimuli and the underestimation of strong stimuli

2. scalar variability found in the Weber-Fechner law.

Weber-Fechner law: $p = k (ln\frac{S}{S_0})$

where the sensation (input of $S_n$) is proportional to the logarithm of the stimulus intensity. $S_0$ is the reference stimulus and S is the change from the reference to the target stimulus.

1. sequential (memory) effects.

2. dynamic range (span in log-units between extremes of stimulus intensities or sensory magnitudes.

Assumption 1: The movement of a Bacillaria colony is driven by information processing. This information processing is neuronal in scope, which means that Bacillaria behavior is both reactive and adaptive.

Assumption 2: Can we distinguish between the behavior of Bacillaria and behavior generated by a network of neurons that produce electrical potentials?

Models of Movement

Investigate the behavior for a series of candidate models for explaining Bacillaria collective movement. Evaluation criterion is that behavioral production occurs in a psychomimetic manner, even though Bacillaria has no nervous system.

While the movement looks "biological", it also looks "psychological", or driven by an adaptive system in a goal-oriented manner. While the goal of the colony is simple (movement through the water column), it requires coordinated filament movements.

Behavioral regulation

Many coordinated collective behaviors can result from the tuning of physical parameters, which has been demonstrated in agent-based models that are analogous to organisms such as the water flea (Daphnia) [7]. This allows us to propose two hypotheses regarding the nature of seemingly intelligent behavior generated by Bacillaria colonies.

Hebbian intelligence: We can apply a version of Hebbian learning where the site of action is filament coupling sites rather than synapses. Hebbian learning can be summarized as "cells that wire together, fire together" [8]. In our case, filaments (or cells) that are adjoint also provide each other with a learning signal. This learning signal provides the basis for a coordinated, sliding motion. Once a pair of filaments learn movements of their neighbors, their independent phase oscillations become entrained to each other [9]. While this seems to happen as a result of mechanical constraints with the onset of movement generation, we ask a more fundamental question: how does this behavior become organized in the development of a Bacillaria colony? The answer is through a form of differential Hebbian learning [10]. Differential Hebbian learning is concerned with the temporal difference between activation and response rather than the causality of the activation itself. The synchronization of filaments may indeed be due to active associative processes.

Pseudo-intelligence: We notice that some seemingly neuronal behaviors are generated by actin filaments along with other components of the colonial cellular matrix. The question might be asked: is the synchronization of movements across filaments (cells) in the colony simply the result of mechanical interactions, or is there a greater degree of autonomous behavior at work? Therefore, while we distinguishing neuronal-generated from non-neuronal-generated behavior. This is similar, at least in principle, to the dilemma of simulated vs. real life-like behavior. In Witkowski and Sinapayen [11], contestants used methods to distinguish between movement trajectories from living organisms (Sharks, Ants, Spiders, and Jaguars) and non-living agents (Boids, Robot Arms, and Artificial Chemical Compounds). The general relation between neuronal and non-neuronal systems can be partitioned as is done in Table 1.

Bacillaria occupy a unique position in this typology: a neuronal instance of behavior generated by non-neuronal processes. We plan to evaluate the potential mechanism behind this type of behavior using deep learning and/or reinforcement learning techniques. This will allow us to propose different neuronal-like cooperative behavioral states generated in our non-neuronal context. Such behavioral states might include feedforward movement generation, density-dependent feedback, light-responsive feedback, and resonant feedback.

Table 1. A demonstration of how autonomy can be partitioned into generated and observed components.

Measurement techniques

Input-Output Relationship [12] and active perception in cells [13], measured using tools such as Signal Detection Theory (SDT) and fold-change detection [14-16].

Magnitude estimation [17] and a route to laws and principles [18-20].

Using a set of light source values, can we establish an input/output curve that determines how much input results in a corresponding degree of output? We can based this on the work of Dvoretskii et.al [21] with developmental Braitenberg Vehicles, where internal model are inferred for a vehilce across the span of development. In this case, we need a similar model for our non-neuronal model. In this case, a connectionist-inspired model of multicellular movement propagation can be modified in a number of ways to uncover the interactions between environmental stimuli and the expression of behavior.

|||| Kuramoto oscillators [22] are used to model synchronized neuronal activity. In our connectionist model, nodes (neuronal units) can act as neuronal oscillators, and the connections are determined by the relative degree of synchronization between neighboring cells.

The network allows us to build an internal model with inputs, outputs, processing units, and interactions. Certain behaviors should conform to a statistical regularity (linear or curvilinear output function).

Psychophysical behaviors produced by a non-neuronal system

There are a number of psychophysical phenomenon that are demonstrated by Bacillaria, and perhaps even Diatoms more generally:

• flow detection.

• resonance detection.

• phototaxis thresholding.

References:

[1] Baluska, F. and Levin, M. (2016). On Having No Head: Cognition throughout Biological Systems. Frontiers in Psychology, doi:10.3389/fpsyg.2016.00902.

[2] Boisseau, R.P., Vogel, D., and Dussutour A. (2016). Habituation in non-neural organisms: evidence from slime moulds. Proceedings of the Royal Society B, 283, 20160446.

[3] Vogel, D. and Dussutour, A. (2016). Direct transfer of learned behaviour via cell fusion in non-neural organisms. Proceedings of the Royal Society B, 28320162382.

[4] Cohn, S.A. (2001). Photo-stimulated effects on diatom motility. Comprehensive Series in Photosciences, 1, 375-401.

[5] Reina, A., Bose, T., Trianni, V., and Marshall, J.A.R. (2018). Psychophysical Laws and the Superorganism. Scientific Reports, 8, 4387.

[6] Lubashevsky, I. (2019). Psychophysical laws as reflection of mental space properties. Physics of Life Reviews, 31, 276–303.

[7] Mach, R. and Schweitzer, F. (2007). Modeling Vortex Swarming in Daphnia. Bulletin of Mathematical Biology, 69(2), 539-562. doi:10.1007/s11538-006-9135-3.

[8] Hebb, D.O. (1949). The Organization of Behavior. New York: Wiley & Sons.

[9] du Bois, N. and Elliott, M.A. (2017). The temporal dynamics involved in object representation updating to predict change. Progress in Brain Research, 236, 269-285.

[10] Cocchi, L., Gollo, L.L., Zalesky, A., Breakspear, M. (2017). Criticality in the Brain: A Synthesis of Neurobiology, Models and Cognition. Progress in Neurobiolology, 158, 132-152. doi:10.1016/ j.pneurobio.2017.07.002.

[11] Witkowski, O. and Sinapayen, L. (2020). Fake Life Recognition Contest. Github, https://github.com/LanaSina/FLR_contest/blob/master/Readme.md

[12] Lan, G. and Tu, Y. (2016). Information Processing in Bacteria: Memory, Computation, and Statistical Physics: a Key Issues Review. Reports on Progress in Physics, 79(5), 052601.

[13] Zakirov, B., Charalambous, G., Aspalter, I.M., Van-Vuuren, K., Mead, T., Kyle Harrington, K., Thuret, R., Ravasz Regan, E., Herbert, S.P., and Bentley, K. (2020). Active Perception during Angiogenesis: Filopodia speed up Notch selection of tip cells in silico and in vivo. bioRxiv, doi:10.1101/2020.08.22.261263.

[14] Adler, M. and Alon, U. (2018). Fold-change detection in biological systems. Current Opinion in Systems Biology, 8, 81-89.

[15] Kerkhof, A.J. and Enquist, B.J. (2009). Multiplicative by nature: why logarithmic transformation is necessary in allometry. Journal of Theoretical Biology, 257, 519–521.

[16] Kello, C.T., Brown, G.D.A., Ferrer-i-Cancho, R., Holden, J.G., Linkenkaer-Hansen, K., Rhodes, T., and Van Orden, G.C. (2010). Scaling laws in cognitive sciences. Trends in Cognitive Science, 14, 223–232.

[17] Poulton, E.C. (1968). The New Psychophysics: six models for magnitude estimation. Psychological Bulletin, 69(1), 1317-1323.

[18] Shepard, R.N. (1987). Toward a Universal Law of Generalization for Psychological Science. Science, 237(4820), 1317-1323.

[19] Chater, N. and Brown, G.D.A. (1999). Scale-invariance as a unifying psychological principle. Cognition, 69, B17–B24.

[20] Chater, N. and Brown, G.D.A. (2008). From universal laws of cognition to specific cognitive models. Cognitive Science, 32, 36–67.

[21] Dvoretskii, S., Gong, Z., Gupta, A., Parent, J., and Alicea, B. (2020). Braitenberg Vehicles as Developmental Neurosimulation. arXiv, 2003.07689.

[22] Breakspear, M., Heitmann, S. and Daffertshofer, A. (2010). Generative models of cortical oscillations: neurobiological implications of the Kuramoto model. Frontiers Human Neuroscience, 4, 190. doi:10.3389/fnhum.2010.00190.

Additional References:

Hart, Y., Goldberg, H., Striem-Amit, E., Mayo, A.E., Noy, L., and Alon, U. (2018). Creative exploration as a scale-invariant search on a meaning landscape. Nature Communications, 9, 5411.

Nosofsky, R.M. (1992). Similarity Scaling and Cognitive Process Models. Annual Reviews in Psychology, 43, 25-53.

Nover, H., Anderson, C.H., and DeAngelis, G.C. (2005). A Logarithmic, Scale-Invariant Representation of Speed in Macaque Middle Temporal Area Accounts for Speed Discrimination Performance. Journal of Neuroscience, 25(43), 10049–10060