By Javier Surasky
Dedicated
to my son and teammate in every goal, Manuel Surasky
A Spanish
version (ES) will be published next Thursday
From human mistakes to the technical production of truth
The 2026
FIFA World Cup is approaching, and in our AI and society series, we want to discuss the Video Assistant
Referee (VAR), presented as a technological solution to a structural problem:
serious refereeing error. Yet far from eliminating disputes, it shifted their
focus: who decides what an image narrates, and how?
Principle
1.1 of the VAR Protocol of the International Football Association Board (IFAB)
states that the system is intended to “assist the referee only in the event of
a ‘clear, obvious and manifest error’ or a ‘serious missed incident’,” and it
limits intervention to four situations: goals, penalty decisions, direct red
cards (not second cautions), and mistaken identity, when the referee cautions
or sends off the wrong player (IFAB, n.d.). The promise was straightforward:
more technology would mean more justice.
That
narrative, however, rests on an assumption rarely made explicit: that referring
error is a perceptual failure. But refereeing is not exhausted by “seeing well;”
it also involves situated interpretation, the application of open-textured
rules, and the exercise of authority. And this is where the problems begin.
Technology
does not remove the human component; it reconfigures it. Collins, Evans, and
Higgins (2016) argue that the incorporation of technologies into officiating,
far from eliminating the human role, redistributes authority and reshapes
decision-making within technical configurations.
In that
sense, VAR does not suppress conflict; it displaces it. Authority no longer
lies exclusively with the on-field referee, visible, identifiable, and
personally accountable, but is distributed across a network of cameras,
operators, protocols, software, and international bodies. The central question
is no longer “What did the referee see?” but rather: how was the image on which
the decision is based produced and understood?
More than
correcting errors, VAR introduces a mutation: from human rights/wrong to a
decision regime grounded in a technical production of evidence that, far from
neutral, is selective, mediated, and, to a significant extent, opaque. And with
that, the problem is no longer only error: it is who produces—and how—what
counts as truth in the game.
VAR as a regime for producing visual truth
VAR is a
device that structures what is taken to be true during the match. It determines
which plays are reviewable, which images are critical, which angles are shown,
and how the decision-making process is articulated; and those determinations
are part of the construction of the very evidence on which the main referee’s
ruling will rely.
VAR’s
legitimacy draws on a longstanding trust in images as proof. Susan Sontag
warned that, although photographs are valued as transparent accounts of
reality, they produce an objectification: “they turn an event or a person into
something that can be possessed,” functioning as a form of “alchemy” rather
than direct access to facts (Sontag, 2003: 81). In this way, the image
necessarily expresses processes of selection, slowing down, interpolation, and
reconstruction guided by those who “capture” it—or, more precisely, generate
it.
Bowker and
Star have shown that every classification rests on categories that “are
historically situated artifacts and, like all artifacts, are learned as part of
membership in communities of practice” (Bowker & Star, 1999: 287). In VAR,
such decisions become visible in minor questions such as which body parts
count, what margin of error is accepted, and which frame fixes the exact
instant to be considered. It is those decisions, and not VAR “itself,” that
determine outcomes, even when the image grounding the decision is mediated
evidence, constructed under technical and normative criteria.
By
“mediated evidence,” we mean evidence that functions as a record of an event,
but whose meaning is shaped by technical and normative processes of selection,
processing, and validation that condition what it signifies.
AI and evidence transformation
The gradual
incorporation of artificial intelligence (AI) systems into officiating
intensifies the logic of displacement through VAR-initiated decentralization.
In the case of semi-automated offside, FIFA (2022: 1) describes a system based
on: (1) “technology capable of automatically generating offside lines in all
areas of the pitch; (2) ball-tracking capability that can automatically
identify the location of the ball and the moment the ball is kicked; (3) a
valid skeletal-tracking system capable of automatically and accurately
identifying skeletal points on multiple players simultaneously; and (4) an
alert system capable of automatically displaying messages when offside
incidents occur.”
Once again,
the core issue is the status attributed to the decision that VAR technology produces.
Offside ceases to be an interpretive matter, contextual and thus debatable, and
becomes an unquestionable technical fact. The decision is no longer the
application of a rule, but the AI’s verification of a situation.
The
“detail” is that, as Hildebrandt warns, computational systems do not merely
represent reality; rather, “Acting ethically concerns making the right
decision, both at the level of individual choice and at the level of designing
the legal, political, and technical choice architectures that frame such
choice” (Hildebrandt, 2020: p. 297). In VAR decisions, especially when assisted by
AI, about how to apply a rule become crystallized and assumed to apply
identically across situations. An open-textured rule is thus transformed into a
binary output: offside / not offside.
As a
result, the space for interpretation narrows in a sport where a player’s
creativity is a positive, admired value, and rule application shifts into a
technological realm beyond the reach of most actors involved. This produces a
second effect: the referee’s authority itself is called into question. The
human referee validates, or becomes part of executing, a decision they did not
make, while the source of that decision becomes diffuse: “the system showed me.”
The
incorporation of AI transforms how decisions are made and how they are
justified, accepted, or contested. The game’s truth shifts from a human view,
with possible human error, to a mere calculation performed by algorithms that
were not only defined by people with the same capacity to err as referees, but
also create a closed system—opaque to those to whom it is applied and
unintelligible to most third parties who watch it: the fans.
The result
is that VAR “was never uncritically received by pundits, managers, or fans.
Critical voices quickly emerged, especially because of VAR’s perceived
ineffectiveness and its ability to construct confusing situations”
(Petersen-Wagner and Lee Ludvigsen, 2023: 870). The opacity inherent in
officiating does not disappear; it simply changes form and becomes better
hidden.
Image, controversy, and a crisis of trust
The 2022
World Cup final between Argentina and France makes the effects of this new
evidence regime especially clear. Many pointed out that before the penalty
awarded for Gonzalo Montiel’s handball, there had been a handball by Dayot
Upamecano. To support their claim, they circulated video images.
What
matters here is the way the conflict shifted toward the image underpinning the
decision: arguments for and against relied on videos of the play from different
angles, speeds, and crops. The dispute concerned which images were valid, which
had been used by VAR during the match, and under what criteria the evidence was
constructed, such that the referee was not called to review the incident.
The
controversy was not about which rule should be applied, but about the process
of producing the visual proof of what “really happened.” Far from closing the
discussion, this keeps it alive over time, reopening it with every new angle or
edit of the footage, and thus, where VAR promised certainty and objectivity,
the very human divergence of opinions about the same event re-emerges in the
face of contradictory evidence.
This case,
notable for occurring in a World Cup final, should not be read as an
exceptional emotional reaction, but as an epistemic symptom. It shows that
trust does not rest on visuality alone and requires answers about control and
the ways media produce the evidence they display. In the era of VAR and AI, the
image ceases to be the endpoint of debate and becomes one of its primary
sources of friction.
The mass
spread of deepfakes through AI has only increased doubts, because altering the
image that VAR will show requires far less technology (and production time)
than creating a good deepfake: removing a frame, shifting a hand or an offside
line by millimeters… is possible in less time than it takes a referee to walk
over and review an incident when summoned.
Empirical evidence and VAR’s unintended effects
Empirical
studies on VAR show that gains in decision accuracy coexist with unintended
effects on authority in the field (Spitz et al., 2021; Zhang et al., 2022;
Archibald et al., 2025).
Spitz et
al. (2021: 147) report that “the predictive odds for making the correct
decision after VAR intervention were significantly higher than for the initial
referee’s decision,” and Zhang et al. (2022) observe improvements in key
decisions in women’s World Cups, while qualitative studies highlight
disciplinary effects in which an interviewed player notes that with VAR “I feel
like I’m constantly being watched [and] I remember being told by the referee
that VAR was watching me” (Archibald et al., 2025: 8). This awareness of
constant observation recalls the Foucauldian panopticon, which induces “a
conscious and permanent state of visibility that assures the automatic
functioning of power” (Foucault, Discipline and Punish: 204).
From this
perspective, greater accuracy is achieved through the presence of the pitch’s
“Big Brother.” The technological effect weighs in before play even begins,
reaffirming that perception, even among players and referees, is more decisive
than the technological contribution considered in isolation.
Image, governance, and power
VAR
operates as an authority capable of defining what occurred in the field from an
institutionally opaque position. As Pielke noted even before its
implementation, “international sports organisations have developed in such a
way that they have less well-developed mechanisms of governance than many
governments, businesses and civil society organisations” (Transparency
International, 2016: 29). Remote technological assistance to referees is part
of that architecture: it has immediate effects, but the technical criteria,
operating protocols, and parameters that structure evidence production are not
transparent: “Private entities develop many sports arbitration algorithms, and
their inner workings are often undisclosed” (Farajpour et al., 2025: 1).
The problem
is compounded because digital audiovisual evidence is inherently unstable: it
depends on specific technical procedures to maintain its integrity. The United
Nations Office on Drugs and Crime warns that “digital evidence poses unique
authentication challenges because of (…) its fragility (i.e., it can easily be
manipulated, altered or damaged)” (UNODC, 2019), while the Scientific Working
Group on Digital Evidence of the United States National Institute of Standards
and Technology (SWGDE) stresses the importance of chains of custody, defining
them as “the chronological documentation of the movement, location, possession,
and disposition of evidence,” and then stating that “maintaining an unbroken
chain of custody is essential to ensure the reliability of digital evidence”
(SWGDE, 2020: 12). Yet none of this is part of VAR’s design, nor have external
audit procedures been established for its systems or for the material they
generate.
Consequently,
VAR’s real power comes from its institutionally protected position against
challenge, redefining the relationship between decision, responsibility, and
legitimacy.
Football,
given its global visibility and symbolic density, makes these
tensions—intrinsic to the introduction of digital technologies into the
governance of human conduct—especially clear, thereby revealing a problem that
extends far beyond sport.
Conclusions
The
incorporation of VAR and AI systems into officiating entails a
deeper transformation: it changes how truth is produced, validated, and
accepted during the match.
The central
problem is how the evidence supporting a decision was constructed. Here, we
have sought to show that VAR does not operate as a neutral mirror of what
occurred, but as a device that produces mediated evidence through prior
normative, technical, and organizational choices that are rarely visible or
open to debate. AI arrives to reinforce that displacement and to heighten
doubts about whether what is being shown is what really happened.
The core
issue of VAR associated with AI is institutional. VAR functions as an authority
capable of influencing decisions critical to a football match, yet it lacks
mechanisms to make its processes transparent, which are naturalized as
technologically objective when they are profoundly human and subjective.
Based on
this diagnosis, some minimum criteria for VAR governance can be proposed:
- First, it is necessary to make
transparent how relevant images are selected, processed, and presented.
- Second, institutional channels
for independent audits of its systems and protocols should be established.
- Third, the traceability of
visual evidence must be ensured so that it can be tracked from capture
through archiving.
- Finally, there must be greater
clarity regarding the role and limits of technical assistance, making
clear that VAR neither does, nor will ever, eliminate the subjectivity
inherent in human decision-making—and the responsibility that follows from
it.
The VAR
case is a dense empirical one that makes the tensions of technologically
mediated decision-making exceptionally visible, because few spaces are as
intensely debated as football, where passion and reason do not always walk
shoulder to shoulder.
References
Archibald,
H.; Mascarenhas, D.R. y Cunningham, I. (2025). “VAR is watching you”:
Professional football players’ experiences and adaptations to VAR
implementation. Managing Sport and Leisure. https://doi.org/10.1080/23750472.2025.2553858
Bowker,
G.C. y Star, S.L. (1999). Sorting things out: Classification and its
consequences. MIT Press.
Collins, H.;
Evans, R. y Higgins, M. (2016). Bad call: Technology’s attack on referees
and umpires and how to fix it. MIT Press.
Farajpour,
R.; Amerinia, M.B. y Pourjavaheri, A. (2025). The Role of Artificial
Intelligence in Arbitration and Legal Challenges Arising from Automated
Decisions in Sports. AI and Tech, Behavioral and Social Sciences, 3(2),
21–28. https://doi.org/10.61838/kman.aitech.3.2.3
Fédération
Internationale de Football Association (FIFA). (2022). Semi-automated offside technology.
Component Testing Overview. https://digitalhub.fifa.com/m/ee666b445da4e29/original/SAOT-Component-Testing-Overview.pdf
Foucault, M. (2002). Vigilar y castigar: Nacimiento de la
prisión. Siglo XXI
Editores.
Hildebrandt,
M. (2020). Law for computer scientists and other folk. Oxford University
Press.
International
Football Association Board (IFAB). (s.f.). VAR protocol. https://www.theifab.com/es/laws/latest/video-assistant-referee-var-protocol/
Petersen-Wagner,
R. y Lee Ludvigsen, J.A. (2023). The video assistant referee (VAR) as
neo-coloniality of power? Fan negative reactions to VAR in the 2018 FIFA Men’s
World Cup, Sport in Society, (26)5, 869-883. Doi: 10.1080/17430437.2022.2070481
Scientific
Working Group on Digital Evidence (SWGDE). (2020). Best practices for
digital forensic video analysis. https://www.swgde.org/wp-content/uploads/2024/04/2024-03-22-SWGDE-Best-Practices-for-Digital-Forensic-Video-Analysis-18-V-001-1.1.pdf
Sontag, S.
(2003). Regarding the pain of others. Farrar, Straus, and Giroux.
Spitz, J.;
Wagemans, J.; Memmert, D.; Williams, A.M. y Helsen, W. F. (2021). Video
assistant referees (VAR): The impact of technology on decision making in
association football referees. Journal of Sports Sciences, 39(2),
147–153. https://doi.org/10.1080/02640414.2020.1809163
Transparency
International (2016). Global Corruption Report: Sport. https://files.transparencycdn.org/images/2016_GCRSport_EN.pdf
United
Nations Office on Drugs and Crime (UNODC). (2019). E4J university module
series: Cybercrime – Module 4: Digital Evidence. https://www.unodc.org/e4j/en/cybercrime/module-4/key-issues/digital-evidence.html
Zhang, Y.,
et al. (2022). The effect of the video assistant referee (VAR) on referees’
decisions at FIFA Women’s World Cups. Frontiers in Psychology, 13,
984367. https://doi.org/10.3389/fpsyg.2022.984367