The good of any one individual is of no more importance, from the point of view of the Universe, than the good of any other.
Henry Sidgwick.
Abstract
Widely accepted theories in modern cosmology say that spacetime is probably infinite. This raises the question how to define a social welfare order (SWO) for an infinite population of people dispersed throughout time and space. Any such SWO should be Lorentz invariant: it should yield the same value independent of the position and velocity of the social observer. I define and axiomatically characterize spatiotemporal Cesàro average utilitarian SWOs as a solution to this problem.
Similar content being viewed by others
Notes
See Section 3 of Pivato (2022) for more detailed summaries of the earlier papers in this list.
A Poisson point process is a random process that generates a random scatter of points in space. A good way to visualize it is as a scatter of raindrops falling on a pavement.
Completeness is not actually required; it just makes the theorem statement slightly shorter.
The elements of \({{\mathcal {N}}}\) are just abstract “labels”; they contain no information about personal identity. To the extent that personal identity is meaningful or ethically relevant, it is encoded in the elements of \({{\mathcal {X}}}\).
Indeed, this is true if \({{\mathcal {X}}}\) is any locally compact metric space –e.g. \({{\mathbb {R}}}^N\).
If \({{\mathcal {X}}}\) is locally compact, then \({{\mathfrak {B}}}({{\mathcal {X}}})=\{\)all locally compact subsets of \({{\mathcal {X}}}\}\). But we don’t use this fact.
u is a “universal utility function”, which evaluates the lifetime utility of every possible life outcome for every possible person. The well-known difficulties in defining such a function are compounded by the fact that future civilizations may contain artificially sentient machines, genetically engineered animals of near-human intelligence, and technologically enhanced “post-humans”, in addition to homo sapiens And the universe may also contain other intelligent species. But these issues are beyond the scope of this paper.
See d’Aspremont and Gevers (2002) for background.
Papers on infinite-population social welfare such as Lauwers (1998) call this a bounded permutation.
This is often called Monotonicity. The term loose Pareto is due to Lauwers and Vallentyne (2004).
Recall the density \(\delta ({{\mathcal {M}}})\) from formula (9). If \(\delta ({{\mathcal {M}}})\) exists, then \({\underline{\delta }}({{\mathcal {M}}})=\delta ({{\mathcal {M}}})\). But the liminf in formula (10) always exists, whereas the limit in formula (9) sometimes does not exist. Thus, \({\underline{\delta }}({{\mathcal {M}}})\) can be seen as a robust generalization of \(\delta ({{\mathcal {M}}})\).
Since our policy choices can only affect our future light cone, it might seem that we could just ignore everyone outside this light cone in our social welfare evaluations. But this assumes an infinite-population SWO for which the population inside any light cone is separable from the population outside of it (a form of separability distinct from Axiom A3). And this SWO must still be both anonymous and Lorentz-invariant. Thus, this paper could be read with \({{\mathcal {O}}}\) interpreted as “the future light cone” rather than “the universe”.
Confusingly, Gale himself calls this relation overtaking, and reserves the term catching up for the slightly weaker condition that \(\liminf _{N{\rightarrow }{\infty }}\sum _{n=1}^N \Big ( u({{\widetilde{x}}}_n)-u({{\widetilde{y}}}_n)\Big )\geqslant 0\).
In the case \({{\mathcal {X}}}={{\mathbb {R}}}\), these are similar to what Khan and Stinchcombe (2023, §3.7) refer to as ergodic sequences.
This is the only place in this paper where the completeness of \(({{\mathcal {X}}},d)\) is used.
The italicised stipulation is important, even in Newtonian mechanics. The kinetic energy of a particle depends nonlinearly on the reference frame of the observer. Thus, the kinetic energy change needed to go from velocity \({{\textbf{v}}}\) to velocity \({{\textbf{w}}}\) also depends on the reference frame of the observer.
For instance, if \({{\mathcal {G}}}\) is a compact topological group, then it always has a right-invariant metric. In particular, if \({{\mathcal {G}}}\) is a compact Lie group, then it has a right-invariant Riemann metric.
This is true if \(\Phi\) is a homogeneous reference system induced by a Lie group acting differentiably on \({{\mathcal {S}}}={{\mathbb {R}}}^N\). In particular, it is true for the Galilean and Relativistic reference systems in Appendix B.
References
Arntzenius F (2014) Utilitarianism, decision theory and eternity. Philos Perspect 28:31–58
Asheim G (2010) Intergenerational equity. Annu Rev Econ 2(1):197–222
Asheim G, Tungodden B (2004) Resolving distributional conflicts between generations. Econ Theory 24(1):221–230
Asheim G, Zuber S (2013) A complete and strongly anonymous leximin relation on infinite streams. Soc Choice Welf 41(4):819–834
Asheim G, Zuber S (2016) Evaluating intergenerational risks. J Math Econ 65:104–117
Asheim G, Bossert W, Sprumont Y, Suzumura K (2010) Infinite-horizon choice functions. Econ Theory 43(1):1–21
Asheim G, d’Aspremont C, Banerjee K (2010) Generalized time-invariant overtaking. J Math Econ 46(4):519–533
Asheim GB, Kamaga K, Zuber S (2022a) Infinite population utilitarian criteria, CESifo Working Paper No. 9576
Asheim GB, Kamaga K, Zuber S (2022) Maximal sensitivity under strong anonymity. J Math Econ 103:102768
Askell A (2018) Pareto principles in infinite ethics. Ph.D. thesis, New York University
Atsumi H (1965) Neoclassical growth and the efficient program of capital accumulation. Rev Econ Stud 32:127–136
Banerjee K (2006) On the extension of the utilitarian and Suppes-Sen social welfare relations to infinite utility streams. Soc Choice Welf 27(2):327–339
Barrow JD, Davies P, Harper CL (eds) (2004) Science and ultimate reality: from quantum to cosmos. Cambridge University Press, Cambridge, UK
Basu K, Mitra T (2003) Aggregating infinite utility streams with intergenerational equity: the impossibility of being Paretian. Econometrica 71(5):1557–1563
Basu K, Mitra T (2007a) On the existence of Paretian social welfare relations for infinite utility streams with extended anonymity. In: Roemer and Suzumura (2007), pp. 85–100
Basu K, Mitra T (2007b) Possibility theorems for aggregating infinite utility streams equitably. In: Roemer and Suzumura (2007), pp. 69–84
Basu K, Mitra T (2007) Utilitarianism for infinite utility streams: a new welfare criterion and its axiomatic characterization. J Econ Theory 133(1):350–373
Bostrom N (2011) Infinite ethics. Anal Metaphys 10:9–59
Brock WA (1970) An axiomatic basis for the Ramsey-Weizsäcker overtaking criterion. Econometrica 38(6):927–929
Chichilnisky G (1996) An axiomatic approach to sustainable development. Soc Choice Welf 13(2):231–257
Chichilnisky G (1997) What is sustainable development? Land Econ 73(4):467–91
Chichilnisky G, Heal G (1997) Social choice with infinite populations: construction of a rule and impossibility results. Soc Choice Welf 14(2):303–318
Cowen T (2007) Caring about the distant future: why it matters and what it means. Univ Chic Law Rev 74:5–40
d’Aspremont C, Gevers L (2002) Social welfare functionals and interpersonal comparability. In: Arrow K, Sen A, Suzumura K (eds) Handbook of social choice and welfare, vol 1. Elsevier, pp 459–541
Diamond PA (1965) The evaluation of infinite utility streams. Econometrica 33(2):170–177
Dorr C, Arntzenius F (2017) Self-locating priors and cosmological measures. In: Barrow JD, Silk J, Chamcham K, Saunders S (Eds) Philosophy of Cosmology. Cambridge University Press, Ch. 20, pp 396–428
Dubey RS (2011) Fleurbaey-Michel conjecture on equitable weak Paretian social welfare order. J Math Econ 47(4–5):434–439
Dubey RS, Mitra T (2011) On equitable social welfare functions satisfying the weak Pareto axiom: a complete characterization. Int J Econ Theory 7(3):231–250
Dubey RS, Mitra T (2014) Combining monotonicity and strong equity: construction and representation of orders on infinite utility streams. Soc Choice Welf 43(3):591–602
Ellis GFR (2007) Issues in the philosophy of cosmology. In: Butterfield J, Earman J (eds) Handbook of the Philosophy of Physics. Elsevier, pp 1183–1286
Ellis GFR, Brundrit GB (1979) Life in the infinite universe. Q J R Astron Soc 20:37–41
Fleurbaey M, Michel P (2003) Intertemporal equity and the extension of the Ramsey criterion. J Math Econ 39(7):777–802
Gale D (1967) On optimal development in a multi-sector economy. Rev Econ Stud 34(1):1–18
Garriga J, Vilenkin A (2001) Many worlds in one. Phys Rev D 64(4):043511
Guth AH (2000) Inflation and eternal inflation. Phys Reports 333:555–574
Guth AH (2001) Eternal inflation. Ann New York Acad Sci 950:66–82
Guth AH (2007) Eternal inflation and its implications. J Phys A Math Theor 40(25):6811
Hamkins JD, Montero B (2000) Utilitarianism in infinite worlds. Utilitas 12(1):91–96
Jonsson A (2020) Infinite utility: time, counterparts and ultimate locations, (preprint)
Jonsson A, Peterson M (2020) Consequentialism in infinite worlds. Analysis 80(2):240–248
Jonsson A, Voorneveld M (2015) Utilitarianism on infinite utility streams: Summable differences and finite averages. Econ Theory Bull 3(1):19–31
Jonsson A, Voorneveld M (2018) The limit of discounted utilitarianism. Theor Econ 13(1):19–37
Khan U, Stinchcombe MB (2018) Planning for the long run: programming with patient, Pareto responsive preferences. J Econ Theory 176:444–478
Khan U, Stinchcombe MB (2023) Intergenerational equity and sustainability: a large population approach. Soc Choice Welf. https://doi.org/10.1007/s00355-023-01450-w
Lauwers L (1997) Rawlsian equity and generalised utilitarianism with an infinite population. Econ Theory 9(1):143–150
Lauwers L (1997) Topological aggregation, the case of an infinite population. Soc Choice Welf 14(2):319–332
Lauwers L (1998) Intertemporal objective functions: strong pareto versus anonymity. Math Soc Sci 35(1):37–55
Lauwers L (2010) Ordering infinite utility streams comes at the cost of a non-ramsey set. J Math Econ 46(1):32–37
Lauwers L (2012) Intergenerational equity, efficiency, and constructibility. Econ Theory 49(2):227–242
Lauwers L, Vallentyne P (2004) Infinite utilitarianism: more is always better. Econ Philos 20(2):307–330
Letavaj P, Mišík L, Sleziak M (2015) Extreme points of the set of density measures. J Math Anal Appl 423(2):1150–1165
Li C, Wakker P (2023) A simple and very general axiomatization of average utility maximization for infinite streams, (working paper)
Linde AD (1990) Particle physics and inflationary cosmology. Harwood Academic
Marinacci M (1998) An axiomatic approach to complete patience and time invariance. J Econ Theory 83(1):105–144
Obata N (1988) A note on certain permutation groups in the infinite dimensional rotation group. Nagoya Math J 109:91–107
Petri H (2019) Asymptotic properties of welfare relations. Econ Theory 67(4):853–874
Pivato M (2014) Additive representation of separable preferences over infinite products. Theory Decis 77(1):31–83
Pivato M (2021) Intertemporal choice with continuity constraints. Math Oper Res 46(3):1203–1229
Pivato M (2022) A characterization of Cesàro average utility. J Econ Theory 201:105440
Pivato M (2023) Population ethics in an infinite universe, (working paper)
Rébillé Y (2007) Patience in some non-additive models. J Math Econ 43(6):749–763
Roemer J, Suzumura K (eds) (2007) Intergenerational equity and sustainability. Palgrave Macmillan, Basingstoke
Sakai T (2003) An axiomatic approach to intergenerational equity. Soc Choice Welf 20(1):167–176
Sakai T (2006) Equitable intergenerational preferences on restricted domains. Soc Choice Welf 27(1):41–54
Sakai T (2010) A characterization and an impossibility of finite length anonymity for infinite generations. J Math Econ 46(5):877–883
Sakai T (2010) Intergenerational equity and an explicit construction of welfare criteria. Soc Choice Welf 35(3):393–414
Sakai T (2016) Limit representations of intergenerational equity. Soci Choice Welf 47(2):481–500
Sleziak M, Ziman M (2008) Lévy group and density measures. J Number Theory 128(12):3005–3012
Steinhardt PJ, Turok N (2002) A cyclic model of the universe. Science 296(5572):1436–1439
Svensson L-G (1980) Equity among generations. Econometrica 48(5):1251–1256
Tegmark M (2004) Parallel universes. In: Barrow et al. (2004), Ch. 21, pp 459–491
Vallentyne P, Kagan S (1997) Infinite value and finitely additive value theory. J Philos 94(1):5–26
von Weizsäcker CC (1965) Existence of optimal programs of accumulation for an infinite time horizon. Rev Econ Stud 32:85–104
Wilkinson H (2021) Infinite aggregation: expanded addition. Philos Stud 178(6):1917–1949
Wilkinson H (2023) Infinite aggregation and risk. Australas J Philos 101(2):340–359
Wilkinson H (2023a) Aggregation in an infinite, relativistic universe. Erkenntnis. https://doi.org/10.1007/s10670-022-00650-7
Wilkinson H (2023b) Chaos, add infinitum, (preprint)
Zame WR (2007) Can intergenerational equity be operationalized? Theor Econ 2:187–202
Zuber S, Asheim GB (2012) Justifying social discounting: the rank-discounted utilitarian approach. J Econ Theory 147(4):1572–1601
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that all sources of funding for this research have been disclosed, and there are no conflicts of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This research supported by Labex MME-DII (ANR11-LBX-0023-01) and CHOp (ANR-17-CE26-0003). I am grateful to Geir Asheim, Jean Baccelli, Marc Fleurbaey, Adam Jonsson, Jean-François Laslier, Christian List, and Bill Zwicker for many excellent suggestions.
Appendices
A Proofs
Proof of Lemma 1
Let \(c\in {{\mathcal {S}}}\) and \(\psi \in \Phi\), and suppose \({\varvec{{\mathcal {K}}}}\) is regular with respect c and \(\psi\). Let \(c'\in {{\mathcal {S}}}\) and \(\psi '\in \Phi\); to show that \({\varvec{{\mathcal {K}}}}\) is also regular with respect \(c'\) and \(\psi '\), we shall check the three conditions of the definition.
-
Mortality. Let \(R_1:=d(c,c')\) and \(R_2:=\delta (\psi ,\psi ')\) and let \(R_*:=\max \{R_1,R_2\}\). Thus, for any \(r\in {{\mathbb {R}}}_+\), we have \({{\mathcal {S}}}(c,r)\subseteq {{\mathcal {S}}}(c',r+R_*)\) and \(\Phi (\psi ,r)\subseteq \Phi (\psi ',r+R_*)\), and thus, \({{\mathcal {O}}}(c,\psi ,r)\subseteq {{\mathcal {O}}}(c',\psi ',r+R_*)\). For any \(n\in {{\mathcal {N}}}\), there exists \(r\in {{\mathbb {R}}}_+\) such that \(n\in {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\), hence \({{\mathcal {K}}}_n\subseteq {{\mathcal {O}}}(c,\psi ,r)\), hence \({{\mathcal {K}}}_n\subseteq {{\mathcal {O}}}(c',\psi ',r+R_*)\), hence \(n\in {{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r+R_*)\). Thus, \(\displaystyle {{\mathcal {N}}}=\bigcup _{r=1}^{\infty }{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\), as desired.
-
Local finiteness. Let \(R_*\) be as already defined above. For any \(r\in {{\mathbb {R}}}_+\), we have \({{\mathcal {S}}}(c',r)\subseteq {{\mathcal {S}}}(c,r+R_*)\) and \(\Phi (\psi ',r)\subseteq \Phi (\psi ,r+R_*)\), and thus, \({{\mathcal {O}}}(c',\psi ',r)\subseteq {{\mathcal {O}}}(c,\psi ,r+R_*)\). Thus, \({{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\subseteq {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\), so \(\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|\leqslant \left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\right|\). Thus, \(\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|<{\infty }\) because \(\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\right|<{\infty }\). This holds for all \(r\in {{\mathbb {R}}}_+\).
-
Subexponential growth. As already noted, \(\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|\leqslant \left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\right|\) for all \(r\in {{\mathbb {R}}}_+\). By the same argument, \(\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|\leqslant \left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r+R_*)\right|\) for all \(r\in {{\mathbb {R}}}_+\). Thus, for any \(r,R\in {{\mathbb {R}}}_+\), we have
$$\begin{aligned} \frac{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r+R)\right|}{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|}\leqslant & {} \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R+R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r-R_*)\right|}.\\ \text{ Thus, }\quad \limsup _{r{\rightarrow }{\infty }} \frac{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r+R)\right|}{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|}\leqslant & {} \limsup _{r{\rightarrow }{\infty }} \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R+R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r-R_*)\right|} \\{} & {} \mathop {=}\limits _{\scriptscriptstyle {\mathrm {(*)}}} \limsup _{r'{\rightarrow }{\infty }} \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r'+(R+2\,R_*))\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r')\right|} \quad \mathop {=}\limits _{\scriptscriptstyle {\mathrm {(\dagger )}}}\quad 1, \end{aligned}$$as desired. Here, \((*)\) is by the change of variables \(r':=r-R_*\), and \((\dagger )\) is because \({\varvec{{\mathcal {K}}}}\) satisfies Subexponential growth relative to c and \(\psi\) by hypothesis.
\(\square\)
Lemma A.1
Let \({\varvec{{\mathcal {K}}}}\in {{\mathfrak {K}}}\), let \(c\in {{\mathcal {S}}}\), and let \(\psi \in \Phi\).
-
(a)
For any \(R>0\), we have \(\displaystyle \lim _{r{\rightarrow }{\infty }}\frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R){\setminus } {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \ = \ 0\).
-
(b)
For any \(c'\in {{\mathcal {S}}}\) and \(\psi '\in \Phi\), we have \(\displaystyle \lim _{r{\rightarrow }{\infty }}\frac{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \ = \ 1\).
Proof
-
(a)
This follows from subexponential growth.
-
(b)
As in the proof of Lemma 1, let \(R_*:=\max \{d(c,c'), \ \delta (\psi ,\psi ')\}\). Then for any \(r>R_*\), we have \({{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r-R_*)\subseteq {{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\subseteq {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\), and thus
$$\begin{aligned} \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r-R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \quad \leqslant \quad \frac{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \quad \leqslant \quad \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|}. \end{aligned}$$Thus,
$$\begin{aligned} 1 \ {}= & {} \ \liminf _{r{\rightarrow }{\infty }} \frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r-R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \ \leqslant \ \lim _{r{\rightarrow }{\infty }}\frac{\left|{{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \ \\\leqslant & {} \limsup _{r{\rightarrow }{\infty }}\frac{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\right|}{\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|} \ = \ 1, \end{aligned}$$where the first and last equalities are by the subexponential growth of \({\varvec{{\mathcal {K}}}}\). The claim follows.
\(\square\)
Proof of Lemma 2
The limit (3) is a special case of the limit (4) when u is the indicator function of the set \({{\mathcal {B}}}\) (in which case \(\{u(x_n)\}_{n\in {{\mathcal {N}}}}=\{0,1\}\) is obviously bounded). So it suffices to show that the existence and value of the limit (4) are independent of the choice of c and \(\psi\).
So, let \(u:{{\mathcal {X}}}{{\longrightarrow }}{{\mathbb {R}}}\), let \({{\textbf{x}}}\in \ell ^{\infty }({{\mathcal {N}}},{{\mathcal {X}}})\), and suppose that \(M:= \sup \{|u(x_n)|\); \(n\in {{\mathcal {N}}}\}\) is finite. Let \(c\in {{\mathcal {S}}}\) and \(\psi \in \Phi\), and suppose the limit (4) exists. Let \(c'\in {{\mathcal {S}}}\) and \(\psi '\in \Phi\); we must show that the limit
also exists, and has the same value as (4). For brevity, I shall use the following notation: for any \(r>0\), \({{\mathcal {N}}}(r):= {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\) and \({{\mathcal {N}}}\,'(r):= {{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\). As in the proof of Lemma 1, let \(R_*:=\max \{d(c,c'), \ \delta (\psi ,\psi ')\}\); then for all \(r\in {{\mathbb {R}}}_+\), we have \({{\mathcal {N}}}\,'(r)\subseteq {{\mathcal {N}}}(r+R_*)\) and \({{\mathcal {N}}}(r)\subseteq {{\mathcal {N}}}\,'(r+R_*)\). Thus, for any \(r\in {{\mathbb {R}}}_+\),
where \((*)\) is because \({{\mathcal {N}}}(r-R_*)\subseteq {{\mathcal {N}}}\,'(r)\subseteq {{\mathcal {N}}}(r+R_*)\). Thus,
where \((*)\) is by (A2). Thus,
where \((\dagger )\) is by Lemma A.1(a,b). It follows that the limit (A1) exists if and only if the limit (4) exists, and if they exist, then they have the same value.
Finally, if \(u:{{\mathcal {X}}}{{\longrightarrow }}{{\mathbb {R}}}\) is continuous, and \({{\textbf{x}}}\in \ell ^{\infty }({{\mathcal {N}}},{{\mathcal {X}}})\), then \(\{u(x_n)\}_{n\in {{\mathcal {N}}}}\) is a totally bounded subset of \({{\mathbb {R}}}\), hence bounded, so the previous analysis applies. \(\square\)
Proof of Lemma 3
Let \(\gamma :=\eta '\circ \eta ^{-1}:{{\mathbb {N}}}{{\longrightarrow }}{{\mathbb {N}}}\). Then \(\gamma\) is a bijection because \(\eta\) and \(\eta '\) are bijections. Clearly, \(\eta '=\gamma \circ \eta\). We must show that \(\gamma\) satisfies formula (8).
As in the proof of Lemmas 1 and 2, let \(R_*:=\max \{d(c,c'), \ \delta (\psi ,\psi ')\}\); then for all \(r\in {{\mathbb {R}}}_+\), we have \({{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\subseteq {{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r+R_*)\). Thus,
Next, for all \(m\in {{\mathcal {N}}}\), formula (6) implies that
Likewise, for all \(m\in {{\mathcal {N}}}\) all \(r<r_{c,\psi }(m)\), formula (6) implies that
For any \(N\in {{\mathbb {N}}}\), let \(r_N:=r_{c,\psi }(n)\), where \(n:=\eta ^{-1}(N)\). Finally, let \(N':=\left|{{\mathcal {N}}}_{\psi ',c'}({\varvec{{\mathcal {K}}}},r_N+R_*)\right|\).
Claim 1
For all \(N\in {{\mathbb {N}}}\), \(\gamma [1\ldots N] \subseteq [1\ldots N']\).
Proof
For all \(m\in \eta ^{-1}[1\ldots N]\), we have \(r_{c,\psi }(m)\leqslant r_N\) by formula (6), and thus, \(r_{c',\psi '}(m)\leqslant r_N+R_*\), by inequality (A3). Thus, \(\eta '(m)\leqslant \left|{{\mathcal {N}}}_{\psi ',c'}({\varvec{{\mathcal {K}}}},r_N+R_*)\right|\) by inequality (A4). The claim follows. \(\diamond\; [Claim 1]\)
Meanwhile, for any \(N\in {{\mathbb {N}}}\) and \(\epsilon >0\),
by inequality (A5). Thus,
where \((\diamond )\) is by Claim 1, \((*)\) is by inequality (A6) and \((\dagger )\) is because \({{\mathcal {N}}}_{c',\psi '}({\varvec{{\mathcal {K}}}},r)\subseteq {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r+R_*)\) for all \(r\in {{\mathbb {R}}}_+\). Combining inequalities (A6) and (A7) yields
But \(r_N{\rightarrow }{\infty }\) as \(N{\rightarrow }{\infty }\). Thus, Lemma A.1(a) implies that
Now combine (A8) and (A9) to obtain (8). \(\square\vspace{1em}\)
The proof of the Main Theorem requires some preliminaries. For any subset \({{\mathcal {M}}}\subseteq {{\mathbb {N}}}\), recall that its density \(\delta ({{\mathcal {M}}})\) is defined by formula (9), and \({{\mathfrak {D}}}\) is the collection of all subsets \({{\mathcal {D}}}\subset {{\mathbb {N}}}\) for which \(\delta ({{\mathcal {D}}})\) is well-defined. This collection is closed under complementation (i.e. \({{\mathcal {D}}}\in {{\mathfrak {D}}}\iff {{\mathcal {D}}}^\complement \in {{\mathfrak {D}}}\)) and under disjoint unions (i.e. if \({{\mathcal {A}}},{{\mathcal {B}}}\in {{\mathfrak {D}}}\) and \({{\mathcal {A}}}\cap {{\mathcal {B}}}=\varnothing\), then \({{\mathcal {A}}}\sqcup {{\mathcal {B}}}\in {{\mathfrak {D}}}\)). But it is not a Boolean algebra. However, \(\delta\) is “finitely additive”: for all disjoint \({{\mathcal {A}}},{{\mathcal {B}}}\in {{\mathfrak {D}}}\), we have \(\delta ({{\mathcal {A}}}\sqcup {{\mathcal {B}}})=\delta ({{\mathcal {A}}})+\delta ({{\mathcal {B}}})\). In fact, \(\delta\) can be extended (non-uniquely) to a finitely additive measure on \(\wp ({{\mathbb {N}}})\). Such measures are called density measures (Letavaj et al. 2015; Sleziak and Ziman 2008).
Recall that \({{\mathfrak {B}}}({{\mathcal {X}}}):=\{{{\mathcal {O}}}\cap {{\mathcal {C}}}\); \({{\mathcal {O}}}\subseteq {{\mathcal {X}}}\) open and \({{\mathcal {C}}}\subseteq {{\mathcal {X}}}\) closed\(\}\). A sequence \({{\textbf{x}}}\in {{\mathcal {X}}}^{{\mathbb {N}}}\) is regular if, for any \({{\mathcal {B}}}\in {{\mathfrak {B}}}({{\mathcal {X}}})\), the set \(\{n\in {{\mathbb {N}}}\); \(x_n\in {{\mathcal {B}}}\}\) is an element of \({{\mathfrak {D}}}\).Footnote 17 Let \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) denote the set of all regular sequences in \(\ell ^{\infty }({{\mathbb {N}}},{{\mathcal {X}}})\). For any \({\varvec{{\mathcal {K}}}}\in {{\mathfrak {K}}}\) and any \({{\textbf{x}}}\in \ell ^{\infty }({{\mathcal {N}}},{{\mathcal {X}}})\) with enumeration \({\widetilde{{\textbf{x}}}}\in \ell ^{\infty }({{\mathbb {N}}},{{\mathcal {X}}})\), it is easy to see that \({{\textbf{x}}}\in \textrm{reg}({\varvec{{\mathcal {K}}}})\) if and only if \({\widetilde{{\textbf{x}}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), by comparing formulae (3) and (9).
If \({{\textbf{x}}}\in {{\mathcal {X}}}^{{\mathbb {N}}}\) and \(\gamma :{{\mathbb {N}}}{{\longrightarrow }}{{\mathbb {N}}}\) is a bijection (e.g. a Lévy permutation), then define \(\gamma ({{\textbf{x}}}):=(y_n)_{n=1}^{\infty }\in {{\mathcal {X}}}^{{\mathbb {N}}}\) by setting \(y_n:=x_{\gamma (n)}\) for all \(n\in {{\mathbb {N}}}\). Clearly \(\gamma ({{\textbf{x}}})\in \ell ^{\infty }({{\mathbb {N}}},{{\mathcal {X}}})\) if and only if \({{\textbf{x}}}\in \ell ^{\infty }({{\mathbb {N}}},{{\mathcal {X}}})\). Furthermore, if \(\gamma \in \Gamma\), then \(\gamma ({{\textbf{x}}})\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) if and only if \({{\textbf{x}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) (Obata 1988, Prop. 3.1).
The Main Theorem is a consequence of a representation theorem of Pivato (2022) for preference orders on \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\). Let \(({{\mathcal {X}}},d)\) be a connected metric space, and let \(\,{\widetilde{\succcurlyeq }}\,\) be a preference order on \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\). We define \(\succcurlyeq _*\) on \({{\mathcal {X}}}\) by restricting \(\,{\widetilde{\succcurlyeq }}\,\) to constant sequences in the obvious way. We require \(\,{\widetilde{\succcurlyeq }}\,\) to satisfy the following four axioms.
- A1*.:
-
Let \({{\mathcal {M}}}\subseteq {{\mathbb {N}}}\) be non-null and let \({{\textbf{x}}},{{\textbf{y}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\). Suppose \({{\textbf{x}}}_{{{\mathbb {N}}}\setminus {{\mathcal {M}}}}={{\textbf{y}}}_{{{\mathbb {N}}}\setminus {{\mathcal {M}}}}\), and there exist \(x,y\in {{\mathcal {X}}}\) such that \(x_m=x\) and \(y_m=y\) for all \(m\in {{\mathcal {M}}}\). If \(x\succ _* y\), then \({{\textbf{x}}}\,{\widetilde{\succ }}\,{{\textbf{y}}}\).
- A2*.:
-
Let \({{\textbf{x}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), and let \(\{{{\textbf{x}}}^k\}_{k=1}^{\infty }\) be a sequence in \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) such that \(\displaystyle \lim _{k{\rightarrow }{\infty }} d_{\infty }({{\textbf{x}}}^k,{{\textbf{x}}})=0\). If \({{\textbf{x}}}^k\succcurlyeq {{\textbf{y}}}\) for all \(k\in {{\mathbb {N}}}\), then \({{\textbf{x}}}\succcurlyeq {{\textbf{y}}}\). If \({{\textbf{x}}}^k\preccurlyeq {{\textbf{y}}}\) for all \(k\in {{\mathbb {N}}}\), then \({{\textbf{x}}}\preccurlyeq {{\textbf{y}}}\).
- A3*.:
-
Let \({{\textbf{x}}},{{\textbf{x}}}',{{\textbf{y}}},{{\textbf{y}}}'\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), and suppose there are subsets \({{\mathcal {J}}}\subseteq {{\mathbb {N}}}\), and \({{\mathcal {K}}}={{\mathbb {N}}}\setminus {{\mathcal {J}}}\) such that \({{\textbf{x}}}_{{\mathcal {J}}}= {{\textbf{x}}}'_{{\mathcal {J}}}\) and \({{\textbf{y}}}_{{\mathcal {J}}}= {{\textbf{y}}}'_{{\mathcal {J}}}\), while \({{\textbf{x}}}_{{\mathcal {K}}}= {{\textbf{y}}}_{{\mathcal {K}}}\) and \({{\textbf{x}}}'_{{\mathcal {K}}}= {{\textbf{y}}}'_{{\mathcal {K}}}\). Then \({{\textbf{x}}}\succcurlyeq {{\textbf{y}}}\) if and only if \({{\textbf{x}}}'\succcurlyeq {{\textbf{y}}}'\).
- A4*.:
-
For all \({{\textbf{x}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), and all \(\gamma \in \Gamma\), \({{\textbf{x}}}\,{\widetilde{\approx }}\,\gamma ({{\textbf{x}}})\).
Axioms A1* - A4* are clearly analogous to Axioms A1 - A4 from Sect. 3. We will also need \(\,{\widetilde{\succcurlyeq }}\,\) to satisfy the following richness condition.
- (B):
-
For any countable, totally bounded \({{\mathcal {Y}}}\subseteq {{\mathcal {X}}}\), there exist \(x,z\in {{\mathcal {X}}}\) such that \(x \preccurlyeq _* y \preccurlyeq _* z\) for all \(y\in {{\mathcal {Y}}}\).
For any sequence \({{\textbf{r}}}=(r_1,r_2,r_3,\ldots )\in {{\mathbb {R}}}^{{\mathbb {N}}}\), the Cesàro average is defined
when this limit exists; in this case, we say that \({{\textbf{r}}}\) is Cesàro summable. If \({{\textbf{r}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathbb {R}}})\) then it is easily verified that the limit (A10) exists. Furthermore,
Proposition A.2
Let \(\,{\widetilde{\succcurlyeq }}\,\) be a preference order on \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) satisfying condition (B). Then \(\,{\widetilde{\succcurlyeq }}\,\) satisfies Axioms A1* - A4* if and only if there is a continuous function \(u:{{\mathcal {X}}}{{\longrightarrow }}{{\mathbb {R}}}\) such that, for all \({{\textbf{x}}},{{\textbf{y}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), we have \({{\textbf{x}}}\,{\widetilde{\succcurlyeq }}\,{{\textbf{y}}}\) if and only if \(\displaystyle \mathop {\textrm{CesAve}}\limits _{n\in {{\mathbb {N}}}} \ u(x_n) \geqslant \mathop {\textrm{CesAve}}\limits _{n\in {{\mathbb {N}}}} \ u(y_n)\).
Proof
See Pivato (2022).\(\square\)
Proof of the Main Theorem
Condition (R) and Axiom A5 imply that there is an order \(\,{\widetilde{\succcurlyeq }}\,\) on \(\textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\) such that, for all \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) and \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) in \({{\mathfrak {S}}}\), with enumerations \({\widetilde{{\textbf{x}}}}\) and \({\widetilde{{\textbf{y}}}}\),
Clearly, \(\succcurlyeq\) satisfies A4 if and only if \(\,{\widetilde{\succcurlyeq }}\,\) satisfies A4*; in this case, it does not matter which enumerations of \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) and \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) we use in statement (A12), by Lemma 3. Meanwhile, \(\succcurlyeq\) satisfies A1 - A3 if and only if \(\,{\widetilde{\succcurlyeq }}\,\) satisfies A1* - A3*. Since the metric space \(({{\mathcal {X}}},d)\) is complete, Axiom A2* implies that \(\,{\widetilde{\succcurlyeq }}\,\) satisfies condition (B), by a proposition in Pivato (2022).Footnote 18 Thus, \(\,{\widetilde{\succcurlyeq }}\,\) has the representation described in Proposition A.2.
Now let \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) be a scenario, and let \({\widetilde{{\textbf{x}}}}\) be an enumeration relative to some \(c\in {{\mathcal {S}}}\) and \(\psi \in \Phi\). Recall that \({{\textbf{x}}}\in \textrm{reg}({\varvec{{\mathcal {K}}}})\) if and only if \({\widetilde{{\textbf{x}}}}\in \textrm{reg}({{\mathbb {N}}},{{\mathcal {X}}})\), and in this case, it is easily verified that
(Recall that the left-hand side of (A13) is independent of the choice of c and \(\psi\), by Lemma 2. Also, it does not matter which enumeration of \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) we use on the right-hand side of (A13), by Lemma 3 and statement (A11).) Now let \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) and \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) be two regular scenarios with enumerations \({\widetilde{{\textbf{x}}}}\) and \({\widetilde{{\textbf{y}}}}\). Then
where \((*)\) is by Proposition A.2, and \((\dagger )\) is by equation (A13). \(\square\)
Proof of Proposition 4.1
If \(\succcurlyeq\) is enumeration-based and satisfies Weak Pareto, then statement (11) implies that \({\widetilde{\succcurlyeq }}\) also satisfies Weak Pareto. In other words, for any \({\widetilde{{\textbf{x}}}},{\widetilde{{\textbf{y}}}}\in \ell ^{\infty }({{\mathbb {N}}},{{\mathcal {X}}})\), if \({{\widetilde{x}}}_n\succ _* {{\widetilde{y}}}_n\) for all \(n\in {{\mathbb {N}}}\), then \({\widetilde{{\textbf{x}}}}\ {\widetilde{\succ }} \ {\widetilde{{\textbf{y}}}}\).
By hypothesis, there is a subspace \({{\mathcal {T}}}\subseteq {{\mathcal {S}}}\) and an isometry \(\gamma :{{\mathbb {Z}}}{{\longrightarrow }}{{\mathcal {T}}}\). Let \(\psi \in \Phi\) be a reference frame (a bijection \(\psi :{{\mathcal {O}}}{{\longrightarrow }}{{\mathcal {S}}}\)). Let \({{\mathcal {Z}}}:=\psi ^{-1}({{\mathcal {T}}})\subseteq {{\mathcal {O}}}\). It suffices to consider scenarios where people’s locations are all in \({{\mathcal {Z}}}\). Let \(\varphi :=\gamma ^{-1}\circ \psi :{{\mathcal {Z}}}{{\longrightarrow }}{{\mathbb {Z}}}\) (a bijection). We shall identify each point in \({{\mathcal {Z}}}\) with its \(\varphi\)-image in \({{\mathbb {Z}}}\).
Recall that \({{\mathcal {N}}}\) is the indexing set. For notational convenience, suppose that \({{\mathcal {N}}}={{\mathbb {Z}}}\). Consider the scenarios \((\varvec{{\mathcal {J}}},{{\textbf{w}}})\), \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) and \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) shown in Fig. 3. In the somata \(\varvec{{\mathcal {J}}}\), \({\varvec{{\mathcal {K}}}}\) and \({\varvec{{\mathcal {L}}}}\), each person’s spatiotemporal location is a single point in \({{\mathcal {Z}}}\). In other words, \(\varvec{{\mathcal {J}}}=({{\mathcal {J}}}_n)_{n=-{\infty }}^{\infty }\), \({\varvec{{\mathcal {K}}}}=({{\mathcal {K}}}_n)_{n=-{\infty }}^{\infty }\) and \({\varvec{{\mathcal {L}}}}=({{\mathcal {L}}}_n)_{n=-{\infty }}^{\infty }\), where \({{\mathcal {J}}}_n=\{j_n\}\), \({{\mathcal {K}}}_n=\{k_n\}\) and \({{\mathcal {L}}}_n=\{\ell _n\}\), for all \(n\in {{\mathcal {N}}}\), with \(j_n\), \(k_n\) and \(\ell _n\) being points in \({{\mathcal {Z}}}\), which we identify with points in \({{\mathbb {Z}}}\) via \(\varphi\). For all \(n\in {{\mathcal {N}}}={{\mathbb {Z}}}\), let \(\varphi (j_n):=9\,n\), \(\varphi (k_n):=9\,n-3\) and \(\varphi (\ell _n):=9\,n-6\), as shown in Fig. 3.
By hypothesis, there is a \({{\mathbb {Z}}}\)-indexed sequence \((v_n)_{n=-{\infty }}^{\infty }\) in \({{\mathcal {X}}}\) such that \(v_n \prec _* v_m\) whenever \(n<m\). For notational simplicity, assume without loss of generality that \({{\mathbb {Z}}}\subseteq {{\mathcal {X}}}\) (with the natural ordering) and \(v_n=n\) for all \(n\in {{\mathbb {Z}}}\). For all \(n\in {{\mathcal {N}}}={{\mathbb {Z}}}\), let \(w_n:= 4\,n\), \(x_n:=4\,n-1\) and \(y_n:=4\,n-2\), as shown in Fig. 3.
Now, let \(a:=\gamma (1)\). It is easily verified that \((\varvec{{\mathcal {J}}},{{\textbf{w}}})\) and \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) have unique enumerations \({\widetilde{{\textbf{w}}}}\) and \({\widetilde{{\textbf{x}}}}\) with respect to a and \(\psi\), namely \({\widetilde{{\textbf{w}}}}=(0\), 4, \(-4\), 8, \(-8\), 12, \(-12\), \(\ldots )\) and \({\widetilde{{\textbf{x}}}}=(-1\), 3, \(-5\), 7, \(-9\), 11, \(-13\), \(\ldots )\). Note that \({{\widetilde{x}}}_n={{\widetilde{w}}}_n-1\) for all \(n\in {{\mathbb {N}}}\). Thus, \({\widetilde{{\textbf{w}}}}\ {\widetilde{\succ }} \ {\widetilde{{\textbf{x}}}}\) by Weak Pareto. Since \(\succcurlyeq\) is enumeration-based and basepoint independent, statement (11) implies that \((\varvec{{\mathcal {J}}},{{\textbf{w}}})\succ ({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\).
Next, let \(b:=\gamma (-2)\). Again, \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\) and \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) have unique enumerations \({\widetilde{{\textbf{x}}}}'\) and \({\widetilde{{\textbf{y}}}}\) with respect to b and \(\psi\), namely \({\widetilde{{\textbf{x}}}}'= (-1\), 3, \(-5\), 7, \(-9\), 11, \(-13\), \(\ldots )\) and \({\widetilde{{\textbf{y}}}}=(-2\), 2, \(-6\), 6, \(-10\), 10, \(-14\), \(\ldots )\). Thus, \({\widetilde{{\textbf{x}}}}' \ {\widetilde{\succ }} \ {\widetilde{{\textbf{y}}}}\) by Weak Pareto. Thus, \(({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\succ ({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\).
Finally, let \(z_n:=4\,n+1\) for all \(n\in {{\mathcal {N}}}={{\mathbb {Z}}}\), and consider the scenario \((\varvec{{\mathcal {J}}},{{\textbf{z}}})\). Let \(c:=\gamma (-5)\). Then \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\) and \((\varvec{{\mathcal {J}}},{{\textbf{z}}})\) have unique enumerations \({\widetilde{{\textbf{y}}}}'\) and \({\widetilde{{\textbf{z}}}}\) with respect to c and \(\psi\), namely \({\widetilde{{\textbf{y}}}}'=(-2\), 2, \(-6\), 6, \(-10\), 10, \(-14\), \(\ldots )\) and \({\widetilde{{\textbf{z}}}}=(-3\), 1, \(-7\), 5, \(-11\), 9, \(-15\), \(\ldots )\). Thus, \({\widetilde{{\textbf{y}}}}' \ {\widetilde{\succ }} \ {\widetilde{{\textbf{z}}}}\) by Weak Pareto. Thus, \(({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\succ (\varvec{{\mathcal {J}}},{{\textbf{z}}})\).
Combining the conclusions of the three previous paragraphs yields \((\varvec{{\mathcal {J}}},{{\textbf{w}}})\succ ({\varvec{{\mathcal {K}}}},{{\textbf{x}}})\succ ({\varvec{{\mathcal {L}}}},{{\textbf{y}}})\succ (\varvec{{\mathcal {J}}},{{\textbf{z}}})\), and hence \((\varvec{{\mathcal {J}}},{{\textbf{w}}})\succ (\varvec{{\mathcal {J}}},{{\textbf{z}}})\) by transitivity. But \(z_n > w_n\) for all \(n\in {{\mathcal {N}}}={{\mathbb {Z}}}\). Thus, Weak Pareto says that \((\varvec{{\mathcal {J}}},{{\textbf{z}}})\succ (\varvec{{\mathcal {J}}},{{\textbf{w}}})\), which is a contradiction. \(\square\)
B Examples of reference systems
Physics provides three important reference systems. In all three systems, \({{\mathcal {S}}}={{\mathbb {R}}}^4={{\mathbb {R}}}\times {{\mathbb {R}}}^3\), with the first dimension representing “time” and the last three dimensions representing “space”, as seen by some observer. Meanwhile, \({{\mathcal {O}}}={{\mathbb {R}}}^4\), and the various elements of \(\Phi\) are different ways of assigning spacetime coordinates to \({{\mathcal {O}}}\).
Basic reference system. For any orthonormal basis \({{\mathcal {B}}}=\{{{\textbf{b}}}^1,{{\textbf{b}}}^2,{{\textbf{b}}}^3\}\) of \({{\mathbb {R}}}^3\), with \({{\textbf{b}}}^n=(b^n_1,b^n_2,b^n_3)\) for \(n=1,2,3\), let \(\rho _{{\mathcal {B}}}:{{\mathcal {O}}}={{\mathbb {R}}}^4{{\longrightarrow }}{{\mathbb {R}}}^4={{\mathcal {S}}}\) be the linear transformation with matrix
In other words, \(\rho _{{\mathcal {B}}}\) leaves the time coordinate unchanged, while it applies to the three spatial coordinates a rotation and/or reflection that maps the standard basis of \({{\mathbb {R}}}^3\) to the basis \({{\mathcal {B}}}\). Meanwhile, for any vector \({{\textbf{v}}}\in {{\mathbb {R}}}^4\), define \(\tau _{{\textbf{v}}}:{{\mathcal {O}}}={{\mathbb {R}}}^4{{\longrightarrow }}{{\mathbb {R}}}^4={{\mathcal {S}}}\) by \(\tau _{{\textbf{v}}}({{\textbf{x}}}):={{\textbf{x}}}+{{\textbf{v}}}\) for all \({{\textbf{x}}}\in {{\mathbb {R}}}^4\); this is “spacetime translation” by the displacement vector \({{\textbf{v}}}\). Finally, let \(\Psi _B:=\{\tau _{{\textbf{v}}}\circ \rho _{{\mathcal {B}}}\); \({{\textbf{v}}}\in {{\mathbb {R}}}^4\) and \({{\mathcal {B}}}\) any orthonormal basis of \({{\mathbb {R}}}^3\}\). Then \(({{\mathcal {O}}},{{\mathcal {S}}},\Psi _B)\) is the reference system of Maxwellian electrodynamics. Note that this system has no way of representing motion.
Galilean reference system. For any vector \({{\textbf{v}}}\in {{\mathbb {R}}}^3\) (representing a velocity), define the Galilean boost \(\gamma _{{\textbf{v}}}:{{\mathcal {O}}}={{\mathbb {R}}}^4{{\longrightarrow }}{{\mathbb {R}}}^4={{\mathcal {S}}}\) by setting \(\gamma _{{\textbf{v}}}(t,{{\textbf{x}}}):=(t,{{\textbf{x}}}-t\,{{\textbf{v}}})\) for all \({{\textbf{x}}}\in {{\mathbb {R}}}^3\) and \(t\in {{\mathbb {R}}}\). This represents a coordinate system assigned to four-dimensional spacetime by an observer traveling with velocity \({{\textbf{v}}}\). Let \(\Phi _G=\{\psi \circ \gamma _{{\textbf{v}}}\); \({{\textbf{v}}}\in {{\mathbb {R}}}^3\) and \(\psi \in \Psi _B\}\); then \(({{\mathcal {O}}},{{\mathcal {S}}},\Phi )\) is the reference system of Newtonian mechanics.
Relativistic reference system. Let \({{\mathfrak {c}}}\) be the speed of light. For any vector \({{\textbf{v}}}=(v_1,v_2,v_3)\in {{\mathbb {R}}}^3\) (representing velocity), let \({\left\| {{\textbf{v}}} \right\| _{{}} } :=\sqrt{v_1^2+v_2^2+v_3^2}\) be its norm, and let \({\widehat{{\textbf{v}}}}=({{\widehat{v}}}_1,{{\widehat{v}}}_2,{{\widehat{v}}}_2):={{\textbf{v}}}/{\left\| {{\textbf{v}}} \right\| _{{}} }\) be the unit vector in the same direction as \({{\textbf{v}}}\). If \({\left\| {{\textbf{v}}} \right\| _{{}} } <{{\mathfrak {c}}}\), then let \(\beta :={\left\| {{\textbf{v}}} \right\| _{{}} } /{{\mathfrak {c}}}\) where \({{\mathfrak {c}}}\) is the speed of light, and let \(\alpha :=1/\sqrt{1-\beta ^2}\) be the “Lorentz factor”. We then define the Lorentz boost \(\lambda _{{\textbf{v}}}:{{\mathcal {O}}}={{\mathbb {R}}}^4{{\longrightarrow }}{{\mathbb {R}}}^4={{\mathcal {S}}}\) to be the linear transformation with matrix
This represents the relativistic coordinate system assigned to four-dimensional spacetime by an observer traveling with relative velocity \({{\textbf{v}}}\). Let \(\Phi _R=\{\psi \circ \lambda _{{\textbf{v}}}\); \(\psi \in \Psi _B\) and \({{\textbf{v}}}\in {{\mathbb {R}}}^3\) with \({\left\| {{\textbf{v}}} \right\| _{{}} } <{{\mathfrak {c}}}\}\); then \(({{\mathcal {O}}},{{\mathcal {S}}},\Phi )\) is the reference system of relativistic physics.
Boost reference systems. For the purposes of the present paper, it is not necessary to incorporate rotations or reflections into the reference system \(\Phi\). Open balls around a point in Euclidean space are already invariant under rotations and reflections around that point. So the sets \({{\mathcal {O}}}(c,\psi ,r)\) and \({{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\) defined in formula (1) will be invariant under rotations and reflections whether or not they are specifically built into \(\Phi\). Likewise, it is not necessary to encode spatial translations into \(\Phi\), since applying a spatial translation is equivalent to changing the central location c in formula (1), and Lemmas 1 and 2 already guarantee that our key constructions are independent of the choice of c. Thus, instead of the basic reference system \(\Psi _B\), we could use the trivial reference system \(\Phi =\{\textrm{Id}\}\). Likewise, instead of the Galilean reference system \(\Phi _G\), we could use the “Galilean boost” reference system \(\Phi '_G=\{\gamma _{{\textbf{v}}}\); \({{\textbf{v}}}\in {{\mathbb {R}}}^3\}\), and instead of the relativistic reference system \(\Phi _R\), we could use the “Lorentz boost” reference system \(\Phi '_R=\{\lambda _{{\textbf{v}}}\); \({{\textbf{v}}}\in {{\mathbb {R}}}^3\) and \({\left\| {{\textbf{v}}} \right\| _{{}} } <{{\mathfrak {c}}}\}\).
Homogeneous reference systems. In all of the examples above, the reference system \(\Phi\) has the form \(\Phi = \{\gamma \circ \phi _0\); \(\gamma \in {{\mathcal {G}}}\}\), where \(\phi _0:{{\mathcal {O}}}{{\longrightarrow }}{{\mathcal {S}}}\) is one bijection, and \({{\mathcal {G}}}\) is a group (in fact, a Lie group) of transformations of \({{\mathcal {S}}}\); in this case, we say that \(\Phi\) is a homogeneous reference system. In the abstract vector space example at the start of § 2.1, \({{\mathcal {G}}}\) was the general linear group \({{\mathbb {G}}}{{\mathbb {L}}}^N\). In the basic reference system \(\Psi _B\), \({{\mathcal {G}}}\) was the group of all rigid motions of three-dimensional space, plus time translations. In the Galilean reference system \(\Phi _G\), \({{\mathcal {G}}}\) was the group of Galilean transformations. In the relativistic reference system \(\Phi _R\), \({{\mathcal {G}}}\) was the group of Poincaré transformations. In the two “boost” reference systems \(\Phi '_G\) and \(\Phi '_R\), \({{\mathcal {G}}}\) was a group of boosts. Such homogeneous reference systems are ubiquitous in modern physics: \(\phi _0\) represents one (arbitrary) coordinate system or representation of the physical world, and \({{\mathcal {G}}}\) is the group of symmetries under which physical laws must remain invariant. However, the framework of this paper does not require the reference system to be homogeneous.
Natural metrics. The framework of Sect. 2 also posits a metric \(\delta\) on \(\Phi\). Clearly, if we want this framework to reflect “the point of view of the Universe”, then \(\delta\) should be non-arbitrary and somehow canonical. In particular, its definition should not depend on a particular reference frame. Let us call this a natural metric.
In the two “boost” reference systems introduced above, there are natural metrics defined via kinetic energy. In both reference systems, reference frames are parameterized by velocities (which are themselves specified with respect to some fixed but arbitrary reference frame). We can thus define distance between two reference frames with velocity vectors \({{\textbf{v}}}\) and \({{\textbf{w}}}\) to be the square root of the kinetic energy of a particle of unit rest-mass travelling at velocity \({{\textbf{w}}}\), from the perspective of an observer at velocity \({{\textbf{v}}}\) (or vice versa).Footnote 19 Equivalently, this is the square root of the minimum energy needed to accelerate a unit rest-mass particle from velocity \({{\textbf{v}}}\) to velocity \({{\textbf{w}}}\), from the perspective of an observer at velocity \({{\textbf{v}}}\) (or vice versa). It is easily verified that this defines a natural metric on the boost reference system. (In particular, it does not depend on the reference frame with respect to which the velocity vectors \({{\textbf{v}}}\) and \({{\textbf{w}}}\) were specified in the first place.)
In the Galilean boost reference system \(\Phi '_G\), this kinetic energy metric reduces to \(\delta (\gamma _{{\textbf{v}}},\gamma _{{\textbf{w}}})={\left\| {{\textbf{v}}}-{{\textbf{w}}} \right\| _{{}} } /\sqrt{2}\) for any velocity vectors \({{\textbf{v}}},{{\textbf{w}}}\in {{\mathbb {R}}}^3\), where \({\left\| \cdot \right\| _{{}} }\) is the ordinary Euclidean norm on \({{\mathbb {R}}}^3\). In the Lorentz boost reference system \(\Phi '_R\), the exact formula is quite complicated, since it involves relativistic velocity addition. For simplicity, we will only consider the case of colinear velocities. Let \({{\mathfrak {c}}}\) be the speed of light, let \(v,w\in (-{{\mathfrak {c}}},{{\mathfrak {c}}})\), and suppose that \({{\textbf{v}}}=(v,0,0)\) and \({{\textbf{w}}}=(w,0,0)\) (relative to some arbitrary reference frame). Then the relativistic speed of a \({{\textbf{w}}}\)-velocity particle as seen from the \({{\textbf{v}}}\)-reference frame is
The relativistic kinetic energy of the \({{\textbf{w}}}\)-particle from the \({{\textbf{v}}}\)-reference frame is then
(assuming a unit rest mass). Thus, in this special case of colinear motion,
(The general formula is more complicated.)
Setting aside kinetic energy, many homogeneous reference systems admit a natural metric. Let e be the identity element of \({{\mathcal {G}}}\), and let \(d_{{\mathcal {G}}}\) be a right-invariant metric on \({{\mathcal {G}}}\) —i.e. a metric such that \(d_{{\mathcal {G}}}(\gamma \circ \eta , \zeta \circ \eta ) =d_{{\mathcal {G}}}(\gamma , \zeta )\), for all \(\gamma ,\zeta ,\eta \in {{\mathcal {G}}}\).Footnote 20 Then \(d_{{\mathcal {G}}}\) induces a natural metric \(\delta\) on \(\Phi\) by stipulating that \(\delta (\phi ,\phi '):=\inf \{d_{{\mathcal {G}}}(\gamma ,e)\); \(\gamma \in {{\mathcal {G}}}\) and \(\gamma \circ \phi =\phi '\}\).
C More on regular somata
This appendix contains some observations that provide a “recipe” for constructing a large class of regular somata. We shall suppose that \({{\mathcal {O}}}\) is a topological space, and for all \(r>0\), suppose that \({{\mathcal {O}}}(c,\psi ,r)\) is an open subset of \({{\mathcal {O}}}\).
Lemma C.1
If \({{\mathcal {K}}}_n\) is compact in \({{\mathcal {O}}}\) for all \(n\in {{\mathcal {N}}}\), then \({\varvec{{\mathcal {K}}}}\) satisfies Mortality.
Proof
\({{\mathcal {O}}}(c,\psi ,r)\subseteq {{\mathcal {O}}}(c,\psi ,R)\) for any \(r<R\). Also, \(\bigcup \limits _{r=1}^{\infty }{{\mathcal {O}}}(c,\psi ,r)={{\mathcal {O}}}\) because \(\bigcup \limits _{r=1}^{\infty }{{\mathcal {S}}}(c,r)={{\mathcal {S}}}\) and \(\bigcup \limits _{r=1}^{\infty }\Phi (\psi ,r)=\Phi\). So \(\{{{\mathcal {O}}}(c,\psi ,r)\}_{r=1}^{\infty }\) is an open cover of \({{\mathcal {O}}}\), hence of \({{\mathcal {K}}}_n\). There is a finite subcover of \({{\mathcal {K}}}_n\) by compactness; thus, there is some \(r\in {{\mathbb {N}}}\) such that \({{\mathcal {K}}}_n\subseteq {{\mathcal {O}}}(c,\psi ,r)\), hence \(n\in {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\). This holds for all \(n\in {{\mathcal {N}}}\). \(\square\)
Now suppose \({{\mathcal {O}}}\) is a metric space in which all open balls have compact closures (e.g. \({{\mathcal {O}}}={{\mathbb {R}}}^N\) with the Euclidean metric).
Lemma C.2
For all \(r>0\), suppose \({{\mathcal {O}}}(c,\psi ,r)\) is contained in an open ball in \({{\mathcal {O}}}\). Suppose \({{\mathcal {K}}}_n\cap {{\mathcal {K}}}_m=\varnothing\) whenever \(n\ne m\), and suppose there exists \(\epsilon >0\) such that for all \(n\in {{\mathbb {N}}}\), the set \({{\mathcal {K}}}_n\) contains an \(\epsilon\)-ball. Then \({\varvec{{\mathcal {K}}}}\) satisfies Local finiteness.
Proof
Any closed ball in \({{\mathcal {O}}}\) can be covered with a finite set of \(\epsilon\)-balls, because it is compact. Thus, it can only contain a finite number of disjoint \(\epsilon\)-balls. But \({{\mathcal {O}}}(c,\psi ,r)\) is contained in such a compact ball, so it also can contain only a finite number of disjoint \(\epsilon\)-balls. The elements of \({\varvec{{\mathcal {K}}}}\) are disjoint. Each contains an \(\epsilon\)-ball. So \({{\mathcal {O}}}(c,\psi ,r)\) can only contain a finite number of elements of \({\varvec{{\mathcal {K}}}}\). Thus, \(\left|{{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|\) is finite. \(\square\)
Finally, suppose that \({{\mathcal {O}}}\) is N-dimensional for some \(N\in {{\mathbb {R}}}_+\), meaning that there is some \(K>0\) such that for any \(R>r>0\), a ball of radius \(R\in {{\mathcal {O}}}\) can contain at most \(K\cdot (R/r)^N\) disjoint balls of radius r. (For example, this is true if \({{\mathcal {O}}}={{\mathbb {R}}}^N\).) Furthermore, suppose there is some L and \(M\in {{\mathbb {R}}}_+\) such that for all \(r>0\), the set \({{\mathcal {O}}}(c,\psi ,r)\) is contained in a ball of radius \(L\cdot r^M\).Footnote 21 If every \({{\mathcal {K}}}_n\) contains a ball of radius \(\epsilon\), then
Thus, \(\left| {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|\) grows at most polynomially in r, hence subexponentially. Now suppose that \(\left| {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right|\) exhibits exactly polynomial asymptotic growth —in other words, there are constants \(P\in {{\mathbb {R}}}_+\) and \(H>h>0\) such that \(h\, r^P \leqslant \left| {{\mathcal {N}}}_{c,\psi }({\varvec{{\mathcal {K}}}},r)\right| \leqslant H\,r^P\) for all sufficiently large \(r\in {{\mathbb {R}}}_+\). Then it is easily verified that \({\varvec{{\mathcal {K}}}}\) satisfies Subexponential growth.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Pivato, M. Cesàro average utilitarianism in relativistic spacetime. Soc Choice Welf 61, 733–761 (2023). https://doi.org/10.1007/s00355-023-01470-6
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00355-023-01470-6