Inhibition and excitation shape activity selection: effect of oscillations in a decision-making circuit

Abstract

Decision-making is a complex task and requires adaptive mechanisms that facilitate efficient behaviour. Here, we consider a neural circuit that guides the behaviour of an animal in ongoing binary choice tasks. We adopt an inhibition motif from neural network theory and propose a dynamical system characterized by nonlinear feedback, which links mechanism (the implementation of the neural circuit) and function (increasing reproductive value). A central inhibitory unit influences evidence-integrating excitatory units, which in our terms correspond to motivations competing for selection. We determine the parameter regime where the animal exhibits improved decision-making behaviour, and explain different behavioural outcomes by making the link between bifurcation analysis of the nonlinear neural circuit model and decision-making performance. We find that the animal performs best if it tunes internal parameters of the neural circuit in accordance with the underlying bifurcation structure. In particular, variation of inhibition strength and excitation-over-inhibition ratio have a crucial effect on the decision outcome, by allowing the animal to break decision deadlock and to enter an oscillatory phase that describes its internal motivational state. Our findings indicate that this oscillatory phase may improve the overall performance of the animal in an ongoing foraging task. Our results underpin the importance of an integrated functional and mechanistic study of animal activity selection. Author summaryOrganisms frequently select activities, which relate to economic, social and perceptual decision-making problems. The choices made may have substantial impact on their lives. In foraging decisions, for example, animals aim at reaching a target intake of nutrients; it is generally believed that a balanced diet improves reproductive success, yet little is known about the underlying mechanisms that integrate nutritional needs within the brain. In our study, we address this coupling between physiological states and a decision-making circuit in the context of foraging decisions. We consider a model animal that has the drive to eat or drink. The motivation to select and perform one of these activities (i.e. eating or drinking), is processed in artificial neuronal units that have access to information on how hungry and thirsty the animal is at the point it makes the decision. We show that inhibitory and excitatory mechanisms in the neural circuit shape ongoing binary decisions, and we reveal under which conditions oscillating motivations may improve the overall performance of the animal. Our results indicate that inefficient or pathological decision-making may originate from suboptimal modulation of excitation and inhibition in the neurobiological network.

Comparison of magnitude-sensitive sequential sampling models in a simulation-based study

“…In particular, an inverse relationship between the minimum strength of the cross-inhibition and the quality values of the nest-sites necessary to break decision-deadlock in the two-alternative case with equal options has been found (Pais et al, 2013). Another nonlinear decision making circuit involving interneuronal units embedded in a nutrition-based framework has also been shown to account for the kind of magnitude-sensitivity which is in qualitative agreement with experimental observations in perceptual decision making (Bose et al, 2019b).…”

“…Furthermore, it has been shown that magnitudes also affect the attentionchoice link (Cavanagh et al, 2014;Smith & Krajbich, 2019). In line with an evolutionary perspective on naturalistic decisions (Pirrone et al, 2014), magnitude-sensitive responses to stimuli have also been found in other areas, such as collective behaviour of social insects (Pais et al, 2013;Reina et al, 2017Reina et al, , 2018Bose et al, 2017) and in an ongoing decision making task related to dietary choice (Bose et al, 2019b).…”

“…In particular, appropriate experimental realisations could help in model selection as magnitude-dependent noise may change model dynamics for different models in different ways, and it could increase our understanding of similarities and differences between perceptual and value-based decisions, such as the effect of overall magnitude or value versus magnitude or value difference, and the origin of fluctuations affecting the decision process. In discriminating between competing hypotheses for magnitude-sensitive decision behaviour, whether input noise processing (Brunton et al, 2013;Teodorescu et al, 2016;Bose et al, 2019a), non-linear utility functions (Bogacz et al, 2007;Tajima et al, 2016), non-linear decision dynamics (Pais et al, 2013;Bose et al, 2019b;Roxin & Ledberg, 2008;Wang, 2002;Wong & Wang, 2006), or any other candidate explanation (for example see Kacelnik et al, 2011), further empirical and theoretical work will be required.…”

Magnitude-sensitivity: rethinking decision-making

“…This account links a biologically and neurally plausible mechanism [80][81][82][83][84][85] of decision-making to magnitude-sensitivity; it has also been shown to fit magnitude-sensitive data well [10,11]. It is interesting to note that several models characterised by nonlinear decision dynamics [8,19,[86][87][88][89] under specific parametrisations can approximate a lateral inhibition model [10,11,50] and show a very high degree of model mimicry when simulated or fitted to data [10,24] -see also Box 2 and Box 3. Moreover, this model is naturally extendable to more than two alternatives [2,79] in a biologically plausible way.…”

“…In the case of a perceptual decision such as deciding which of two stimuli is brighter, the magnitude is the overall brightness of the alternatives, so that two stimuli having brightness of 30 cd/m 3 and 60 cd/m 3 have a magnitude of 30 + 60 = 90 cd/m 3 . 'Magnitude-sensitivity' (often also referred to as 'value-sensitivity') refers to the result that performance in decision-making is affected by the magnitude of the alternatives [1,3,[8][9][10][11][12][13][14][15][16][17][18][19][20][21]. In particular, for higher magnitude conditions, decision-makers show responses that are faster and more likely to be random (i.e., more likely to be incorrect when there is a correct alternative) [3,11,22].…”