Every living being, simply by living, imposes a cost on other living beings. This is not a modern ecological insight; the Bhāgavata Purāṇa stated it with startling directness millennia ago, in a verse Nārada speaks to Yudhiṣṭhira:
ahastāni sahastānām
apadāni catuṣ-padām
phalgūni tatra mahatāṃ
jīvo jīvasya jīvanam “Those who are devoid of hands are prey for those who have hands; those devoid of legs are prey for the four-legged. The weak are the subsistence of the strong, and the general rule holds that one living being is food for another.” Śrīmad-Bhāgavatam 1.13.47
Jīvo jīvasya jīvanam — one life is the life of another. There is no diet, however carefully chosen, whose harm is exactly zero. Even the strictest vegetarian ends the lives of plants, displaces insects in the field, consumes water that other creatures needed. The honest question, then, is not the binary one — to harm or not to harm? — for that question has no feasible answer. The honest question is an optimization problem: given that survival has a cost in life, how should that cost be minimised?
The problem, stated properly
Let there be \(k\) available food sources, and let \(x_i \ge 0\) denote the quantity of source \(i\) consumed per unit time. Each source carries a nutritional profile \(n_{ij}\) — the amount of nutrient \(j\) supplied per unit of source \(i\) — and a harm coefficient \(h_i \ge 0\): the expected suffering inflicted per unit consumed. A diet \(x = (x_1, \dots, x_k)\) is feasible if it sustains the body. The problem is then the linear program
\[ \min_{x \,\ge\, 0} \; \sum_{i=1}^{k} h_i\, x_i \qquad \text{subject to} \qquad \sum_{i=1}^{k} n_{ij}\, x_i \;\ge\; N_j, \quad j = 1, \dots, m, \]
where \(N_j\) is the body’s requirement of nutrient \(j\). Two features of this formulation deserve attention. First, the constraint set is non-empty — survival is permitted; the Śāstra never demands self-starvation. Second, the objective is minimisation, not elimination: the optimum \(\sum h_i x_i^\ast\) will be strictly positive. Ahiṃsā, in this reading, is not the impossible claim that one can live harmlessly; it is the discipline of living at the minimum of the harm functional.
Estimating the harm coefficients
The mathematics is trivial; the statistics is not. The entire content of the problem lies in the coefficients \(h_i\), and suffering is only partially observable. Yet we do not need precise values — a linear programme’s minimiser is determined by the relative ordering of the ratios \(h_i / n_{ij}\), so a defensible partial order on the \(h_i\) suffices. And nature provides estimators.
Consider how each food is relinquished by its source. A ripe fruit detaches from the tree by itself — botanists call it abscission: the plant builds a dedicated layer of cells whose function is to let go. The mango is, in a quite literal sense, offered; the tree’s reproductive strategy depends on the fruit being eaten and the seed carried away. Grains are gathered from annual grasses at the natural end of their life cycle. Milk, in the traditional economy, is taken from an animal that continues to live and is cared for. Now compare the other column of the menu: no limb of any animal detaches by itself. Flesh is obtained only by traumatic interruption of a life that was, by every behavioural and neurological measure, strenuously trying to continue. Animals possess nociceptors, centralised nervous systems, stress hormones, and unmistakable avoidance behaviour; a fruit tree exhibits none of these in respect of its fruit. Whatever one’s uncertainty about plant sentience, the difference in harm between a fallen fruit and a slaughtered animal is not a matter of decimal places — it is orders of magnitude, in opposite directions of the scale.
So we may write, with more confidence than we usually have about any estimated parameter,
\[ h_{\text{fruit}} \;\approx\; 0 \;<\; h_{\text{grain}} \;<\; h_{\text{root}} \;\ll\; h_{\text{animal}}. \]
The conclusion is then forced by the structure of the programme itself. Because the harm gap between plant and animal sources dwarfs any difference in their nutritional profiles, the minimiser \(x^\ast\) concentrates its mass entirely on the plant side of the menu: fruits, grains, legumes, vegetables, milk. The solution is also robust — the statistician’s favourite property: perturb the \(h_i\) by any reasonable amount, model plant suffering as generously as you like, and the argmin does not move. A vegetarian diet is not the sentimental choice; it is the stable solution of the harm-minimisation programme under every defensible specification of its parameters.
The constraint set, and who wrote it
It is worth asking why the constraint is \(\sum n_{ij} x_i \ge N_j\) and not “maximise enjoyment.” The Īśopaniṣad opens with precisely this point — consumption is legitimate only up to one’s allotted share:
īśāvāsyam idaṃ sarvaṃ
yat kiñca jagatyāṃ jagat
tena tyaktena bhuñjīthā
mā gṛdhaḥ kasya svid dhanam “Everything animate or inanimate within the universe is controlled and owned by the Lord. One should therefore accept only those things necessary for himself, which are set aside as his quota, and not accept other things, knowing well to whom they belong.” Śrī Īśopaniṣad, Mantra 1
Read mathematically, the mantra fixes the feasible region: take what is needed (tena tyaktena bhuñjīthā — accept your set-aside portion), and do not covet beyond it (mā gṛdhaḥ). Greed is the act of replacing a constraint-satisfaction problem with an unconstrained maximisation — and most of the violence in the food economy enters through exactly that substitution.
And the Bhagavad-gītā specifies, with the brevity of a well-posed boundary condition, which column of the menu is fit to be offered at all:
patraṃ puṣpaṃ phalaṃ toyaṃ
yo me bhaktyā prayacchati
tad ahaṃ bhakty-upahṛtam
aśnāmi prayatātmanaḥ “If one offers Me with love and devotion a leaf, a flower, fruit or water, I will accept it.” Bhagavad-gītā 9.26
A leaf, a flower, a fruit, water: notice that the list is precisely the set of foods whose harm coefficients sit at the bottom of our ordering. The gītā’s menu and the linear programme’s argmin coincide — which should not surprise us. The tradition had solved the optimization problem by inspection, long before anyone wrote down the Lagrangian.
What the model does and does not claim
A model is a lens, not a verdict, and intellectual honesty requires stating its limits. The harm coefficients are estimated, not measured; suffering admits no direct observation, and one should hold the numbers loosely even while trusting the ordering. The linearity is a convenience; real harm may compound. And no optimization argument captures what is, for the practitioner, the heart of the matter — that food prepared and offered with devotion (prasāda) transforms eating from an act of consumption into an act of participation.
But perhaps that is the deepest point of contact between the two ways of thinking. Both the statistician and the śāstra begin by refusing a comfortable illusion: that our living costs nothing. Both then ask the same disciplined question — given the cost, what is the least of it? The Bhāgavatam states the constraint, the Upaniṣad bounds the feasible region, the Gītā exhibits the solution. The mathematics merely checks the arithmetic of a wisdom that was always there: jīvo jīvasya jīvanam is the model, and ahiṃsā is its minimiser.