CSR: why overly simplistic indicators can be misleading

For years, we’ve heard the entic­ing sto­ry that dig­i­tal tech­nolo­gies will save the plan­et. Swap flights for video­con­fer­enc­ing, replace CDs with stream­ing, opti­mise traf­fic with AI—the log­ic seems clear: few­er phys­i­cal resources, more dig­i­tal activ­i­ty, low­er emis­sions. But this is only part of the sto­ry. Dig­i­tal sys­tems may seem intan­gi­ble, but they depend heav­i­ly on mate­r­i­al real­i­ty: chips require rare-earth min­ing, and data cen­tres con­sume vast amounts of water for cool­ing. As these tech­nolo­gies become more pop­u­lar, users also change their behav­iours, cre­at­ing new demands that are hard to mea­sure. Cap­tur­ing the full, sys­temic impact of these tech­nolo­gies is a com­plex challenge.

Nev­er­the­less, tools cap­tur­ing sys­temic impacts are more impor­tant than ever. In France, tech com­pa­nies have shift­ed from resist­ing envi­ron­men­tal reg­u­la­tions to start­ing to vol­un­tar­i­ly inte­grate assess­ment tools, some­times repur­posed beyond their orig­i­nal scope. These are becom­ing key man­age­ment instru­ments that are sup­posed to influ­ence strat­e­gy, invest­ment deci­sions, and com­pet­i­tive­ness. In our paper, we tack­le the dual chal­lenge of accu­rate­ly assess­ing the envi­ron­men­tal impacts of dig­i­tal technologies—especially their com­plex sys­temic effects—and ensur­ing that these eval­u­a­tion tools are deeply inte­grat­ed with­in organ­i­sa­tions to dri­ve actu­al transformation.

In this paper, we dis­cuss exist­ing lit­er­a­ture, inter­na­tion­al stan­dards (such as ISO 14040¹ and ITU L.1410)², and pro­vide exam­ples about how the envi­ron­men­tal impacts of dig­i­tal tech­nolo­gies are cur­rent­ly measured.

We find that we’re often assess­ing dig­i­tal impacts too nar­row­ly. Eval­u­a­tion tools too often focus on first-order (direct) effects—e.g. direct life‑cycle impacts of devices, extrac­tion, man­u­fac­ture, trans­port, use, dis­pos­al of mate­ri­als for com­po­nents, and envi­ron­men­tal cost of run­ning data centres—rather than socio-eco­nom­ic con­sid­er­a­tions. Such is the case, for exam­ple, of the wide­ly used green­house gas emis­sions bal­ance sheet (Bilan Car­bone ©) or of direct life-cycle assess­ment tools.

This is rel­e­vant for car­bon report­ing, but ignores the knock-on effects of the tech­nol­o­gy, such as feed­back loops like rebound effects, infra­struc­ture build‑out, and behav­iour­al shifts. Take the roll-out of 5G³ as an exam­ple. At first glance, 5G is a win­ner for envi­ron­men­tal gain. It trans­mits more giga­bits per unit of ener­gy, hence using 5G should reduce the ener­gy used to trans­mit infor­ma­tion. How­ev­er, 5G effec­tive­ly puts ultra-high-def­i­n­i­tion stream­ing in people’s pock­ets, giv­ing access to data-heavy mate­r­i­al all day, any­where, at cut prices. This ease of access entails an increase in demand, adding to the weight of the tech­nol­o­gy on green­house gas­es. More­over, with more demand comes more infra­struc­ture and more hard­ware, and, as it fol­lows, more pre­cious met­als, ener­gy, car­bon emis­sions, and human health dam­ages through­out the whole man­u­fac­tur­ing process.

His­tor­i­cal par­al­lels exist. While pro­vid­ing huge effi­cien­cy gains, mech­a­ni­sa­tion, elec­tri­fi­ca­tion, and automa­tion all increased total ener­gy con­sump­tion⁴ and resource use in the long run. The dig­i­tal sec­tor appears to fol­low the same trend. What we pro­pose to call the “sys­temic effects” (more often referred to in lit­er­a­ture as sec­ond-order and third-order effects, regroup­ing indi­rect changes in behav­iour, demand growth, and macro-eco­nom­ic trans­for­ma­tions) are usu­al­ly not cap­tured by assess­ment tools, per our analysis.

Some tools exist to cap­ture sec­ond-order and third-order effects, but these can be impre­cise and biased. This leads to wild­ly opti­mistic assess­ments like Glob­al e‑Sustainability Ini­tia­tive (GeSI)’s⁵ pro­jec­tion that ICT could cut glob­al GHG emis­sions by 20% by 2030, or GSMA’s⁶ claim that mobile net­works “avoid­ed” ten times their direct emis­sions⁷. Oth­er tools can be data hun­gry and dif­fi­cult to apply to an organ­i­sa­tion­al lev­el, like the con­se­quen­tial life‑cycle assess­ment (CLCA)⁸, which can pro­duce sce­nario ranges rather than pre­cise fig­ures. These tools, more rel­e­vant for sys­temic analy­sis but also more com­plex, have lit­tle uptake among organisations.

In short, mak­ing eval­u­a­tion meth­ods more rig­or­ous and accu­rate can also make them so com­plex that they no longer help peo­ple learn or dri­ve change.

Even the best method is pow­er­less if it stays locked with spe­cial­ists. Many organ­i­sa­tions out­source dig­i­tal foot­print­ing, receiv­ing only a report in return. How­ev­er, learn­ing and the poten­tial shift in mind­set hap­pen in the course of the process itself, not just with pro­duc­ing the final fig­ures. Research shows that tools can spark change only when embed­ded in rou­tines, dis­cussed across teams, and revis­it­ed over time. Faced with missed envi­ron­men­tal goals, some firms engage in re-oper­a­tional­i­sa­tion; low­er­ing their tar­gets rather than chang­ing strat­e­gy, sus­tain­ing “busi­ness as usu­al” under a green­er label.

Our analy­sis sug­gests that the envi­ron­men­tal dynam­ics of dig­i­tal­i­sa­tion require a shift from a sta­t­ic, attri­bu­tion­al approach to a more dynam­ic, sys­temic and con­se­quen­tial one. From iso­lat­ed report­ing to col­lab­o­ra­tive learn­ing and from com­fort­ing nar­ra­tives to evi­dence-based real­ism. We pro­pose using open­ness as a lever to achieve this. Open approach­es make results more trans­par­ent and eas­i­er to relate to their under­ly­ing assump­tions. They let organ­i­sa­tions appre­hend these meth­ods, even with­out large bud­gets. Final­ly, we need real-world research on how such meth­ods are applied in prac­tice, so that we can under­stand how they actu­al­ly help organ­i­sa­tions learn and transform.