“A
Languageless Love”: Untranslatability in Syed Mujtaba Ali’s Shabnam
and the Limits of AI-Mediated Literary Pedagogy
Mohammad Jashim Uddin,
Associate Professor,
Department of English,
Northern University Bangladesh,
Dhaka, Bangladesh.
Abstract: Syed Mujtaba Ali’s
Bengali novel Shabnam (1960)
has attracted comparatively little sustained criticism in English, and none, to
this writer’s knowledge, that reads it alongside the computational turn in
literary study. This article argues that Shabnam
advances a theory of meaning fundamentally incompatible with the assumptions
underwriting machine translation and large language models. Set in Kabul during
the collapse of Amanullah Khan’s reformist state, the novel layers Persian,
Arabic, Urdu, Pashto and English through Bengali script, and repeatedly stages
comprehension as something that either exceeds or falls short of semantic
transfer. Two episodes are decisive: the narrator’s memory of carrying two
Qur’anic words since childhood without ever fully grasping them, and an
embedded tale of “languageless love” in which a Bedouin woman crosses the world
equipped with two English words. Reading these against theories of cultural
translation and untranslatability, and against recent critiques of linguistic
inequality in natural language processing, the article proposes that Shabnam functions as a diagnostic
text for the AI-mediated classroom. It concludes with a pedagogical framework
for Bangladeshi undergraduate teaching that treats machine failure as its
primary instrument.
Keywords: Syed Mujtaba Ali; untranslatability; cultural
translation; artificial intelligence and pedagogy; Bengali literature
I. The Last Two Words
Syed Mujtaba Ali’s Shabnam closes not with its
heroine but with a borrowed story. The Bengali narrator, waiting in a Kabul
emptied by civil war, remembers a tale he read at ten or twelve in “a slim
Bengali book” he was never able to find again (Ali 269). An Englishman is
captured by a Bedouin band; the chief entrusts his daughter with feeding the
prisoner; and between the two, Mujtaba Ali writes, “a languageless love arises”
— bhāṣāhīnpraṇay hay du’janāte (269). The Englishman escapes, leaving behind
exactly two words: his name and the name of his city. The girl follows him
across the sea. On every ship she says only “Tom” — “London.” A captain takes
pity and sets her down in the city itself, and there, the narrator notes, she
walks through “that vast human ocean” having learned “not one further letter of
English,” answering every question put to her with a wan smile and the same two
syllables (270). And she finds him. They embrace in the middle of the main
road. The narrator draws the only conclusion he can bear: “Won’t Shabnam
come just like that, one day?” The novel’s final line is not in Bengali at all.
It is Persian: tamāmshud — it is finished (270).
It is a strange way to end a love story, and stranger
still for a writer of Mujtaba Ali’s formidable learning. He had studied at
Santiniketan under Rabindranath Tagore, taken a doctorate at Bonn, read at
al-Azhar, and was reputed to command a dozen languages. Yet the last thing his
most famous novel does is stake everything on a woman who knows two words. This
article takes that ending seriously as a theoretical proposition rather than a
sentimental flourish. Shabnam, I argue, is organised around a claim
about meaning that current language technologies are constitutionally unable to
model: that what passes between people is routinely detached from what they
understand, and that the residue which survives translation may be precisely
the part that was never comprehended in the first place.
The argument matters beyond Mujtaba Ali. Bengali
literature enters the English-medium classrooms of Bangladesh increasingly
through machine-assisted paths — automatic translation, generative summary,
AI-produced study notes. Those systems are engineered on the opposite premise
from the one Shabnam advances: that meaning is recoverable, transferable
and, given enough parameters, complete. Reading Shabnam is therefore not
merely an occasion to complain about technology. The novel supplies a positive
account of what the machine is missing, and in doing so it becomes unusually
valuable as a teaching text at exactly the moment its pedagogy is most under
pressure.
II. A Bengali Novel of Afghanistan
Mujtaba Ali lived in Kabul from roughly 1927 to 1929,
teaching in connection with the Afghan Ministry of Education during the
reformist reign of Amanullah Khan, and left as the Saqqawist uprising brought
that order down. He wrote the period twice. Deshe Bideshe (1948), his celebrated
debut, records it as travelogue. Shabnam (1960) returns to the same
months as romance. The pairing is itself instructive: the second book is not a
novelisation of the first so much as a test of what the fictional frame can
hold that reportage could not, and the most conspicuous thing it holds is a
woman.
The novel opens on the register of comic scepticism that
made Mujtaba Ali a household name in Bengal: “King Amanullah must have lost his
mind. Why else would he go arranging a ball-dance in a country as cussed and
bigoted as Afghanistan?” (Ali 111). The joke does a great deal of work. It
establishes the narrator as an insider-outsider, licensed to mock the Afghan
court and the Afghan clergy in the same breath; it foregrounds Amanullah’s programme
of forced cosmopolitan modernity; and, crucially, it does so through a
transliterated English compound, bal-ḍyāns, sitting inside a Bengali sentence
about Islamic orthodoxy. Three linguistic worlds are stacked in a single clause
before the plot has begun.
Shabnam
herself first appears in a Kabul tea-room, in a scene of considerable political
density that the novel refuses to underline. Half a dozen Kabuli young women
have lowered the veils from their heads and sit around a table “chattering”
(Ali 121) — the visible face of Amanullah’s reforms. Shabnam is at the
counter, berating a waiter at speed about the absence of pastry. She is
introduced, that is, through voice and grievance rather than through
description, and through a European loanword. When she and the narrator finally
reach an understanding, it is she who proposes: “I have decided we will marry.
And you?” and then, without waiting, “I know your eyes. You need not say
anything. This time I have won” (Ali 156). She sets the date for the same
evening.
This matters for the reading that follows, because it is
tempting to file Shabnam among the idealised beloveds of the Persianate
romance tradition on which the novel draws so heavily. The text will not quite
allow it. She decides, she initiates, she wins; the narrator is the one who
cannot get his mouth open. What she is denied is not agency but futurity. It is
worth registering, too, how the novel handles the marriage itself: the
negotiation is conducted almost entirely by her family, whose patriarch tells
the narrator flatly, “You are a foreigner — you do not know the conditions of
this country” (Ali 191), and the guest list runs to the narrator’s servant
Abdur Rahman, carried over from Deshe Bideshe. Cosmopolitan romance is
nested inside a thoroughly local machinery of kinship and obligation. Sheldon
Pollock’s account of the cosmopolitan and the vernacular as mutually
constituting rather than opposed orders is apposite here: Mujtaba Ali’s Kabul
is not a space where the local dissolves into the universal, but one where the
two are continuously negotiated, sentence by sentence.
A word is needed about the Orientalism question,
since a Bengali writing Afghanistan invites it. Edward Said’s framework,
developed against a European archive, transfers awkwardly to intra-Asian
representation, and Shabnam both invites and frustrates the charge. The
narrator retails Afghan “bigotry” for Bengali laughter; he also, in one of the
novel’s most quietly radical passages, has Shabnam interrogate him in
detail about Balkh, Bamiyan, Gandhara, Peshawar and Nalanda — not as exotica
but because “they lie on the road” to her father-in-law’s home in East Bengal
(Ali 251). The itinerary is Xuanzang’s, reversed. Afghanistan and Bengal are
placed on a single Buddhist-Persianate continuum along which a Kabuli bride
might simply travel. Whatever else this is, it is not the Orient constructed as
a static object of European knowledge; it is a claim of shared ground made in
the idiom of a woman planning a journey she will not take.
III. The Polyglot Page
Shabnam is
written in a Bengali that is constantly hosting other languages. Persian and
Urdu couplets run through the courtship, sometimes glossed, sometimes not.
Qur’anic Arabic appears in Bengali script. English words — ball-dance, pastry,
and later the two words of the closing tale — arrive as transliterations. The
preceding piece in the same volume ends with English epitaph verse set in Roman
type on a Bengali page. Even the novel’s dedication, to Rajshekhar Basu, is
written in the formal sādhu register while the narrative itself proceeds in the
colloquial calit, so that the reader crosses a diglossic boundary before
reaching the first sentence.
This is not ornament. It is the novel’s way of making a
claim about how understanding actually operates. Consider the narrator’s
account of speaking to Abdur Rahman: “I spoke in my mother tongue, Bengali. Abdur
Rahman was a plain-dealing man; he understood without understanding Bengali”
(Ali 111). The sentence is built on a contradiction it declines to resolve.
Comprehension is asserted and simultaneously denied at the level of language.
What passes between the two men is real, and it is not semantic.
Talal Asad’s critique of “cultural translation” in
British social anthropology is useful for naming what is at stake. Asad’s
objection was that the anthropologist claims to render another society’s
implicit meanings explicit, and in doing so silently substitutes his own
coherence for theirs. Mujtaba Ali’s narrator is exactly the figure Asad
describes — the learned outsider who can gloss everything — and the novel keeps
undercutting him. He is the man who knows a dozen languages and cannot say yes
to a proposal. Homi Bhabha’s formulation of translation as the site where
cultural authority is negotiated rather than transferred describes the same
structure from the other side: what the polyglot page enacts is not mastery
over several tongues but the permanent incompleteness of passing between them.
IV. Two Words: The Novel’s Theory of Incomplete Meaning
The theory becomes explicit in a short passage near the
novel’s close, and it is worth quoting at length because everything I want to
argue depends on it. Shabnam has gone; her Qur’an is still on the table.
The narrator opens it at random, in the manner of those seeking an augury,
though he insists he seeks none. He reads a verse of Arabic. And then, instead
of interpreting it, he remembers:
My brothers were at school; I had not reached that age.
At noon my mother would dress me in a broad-bordered dhuti, a lace-sleeved
kurta she had woven herself, and a gold-worked cap, and sit me down before her
to read the Qur’an. … What stayed with me was only two words — fazl and rahmat,
overflowing bounty and mercy. At the time I understood neither their meaning
nor anything else. Have I fully understood them even now? (Ali 269)
Several things happen at once. The child is a Bengali Muslim
boy in a dhuti, a garment culturally coded Hindu, learning Arabic scripture
from his mother in a domestic Bengali setting: four cultural systems in one
image, none of them in conflict, none of them translated into the others. And
the pedagogical scene is a scene of non-comprehension that nonetheless
transmits. Two words survive fifty years. They survive as sound, as maternal
association, as the residue of an afternoon — and the narrator’s closing
question concedes that the semantic content has still not been fully acquired.
Meaning here is not a thing possessed but a debt carried.
The Bedouin tale that follows, and closes the book, is
the same structure told again in the register of romance. Two words, again. Not
understood, again — the girl learns “not one further letter of English.” And,
again, sufficient: the two words carry her across an ocean and through a city
of millions to the person they name. Placed side by side, the Qur’anic memory
and the Bedouin story make an argument. What travels is not what is understood.
The untranslated fragment is load-bearing precisely because it has not been
dissolved into an equivalent.
Emily Apter’s Against World Literature supplies
the term for this: untranslatability, not as a failure state to be regretted
but as a positive value that resists the frictionless circulation of texts
through a global English. Lawrence Venuti’s account of domestication describes
what the alternative looks like in practice — the fluent target-language
rendering that erases the labour and the foreignness of its own production.
Read through these, tamāmshud is not a decorative flourish. It is a refusal.
The novel ends in a language it declines to translate for a readership it knows
cannot fully read it, having spent two hundred pages arguing that this is how
meaning has always worked.
V. What the Machine Recovers, and What It Cannot
Set this against the operating assumptions of
contemporary language technology. A neural translation system is trained to
produce, for a given input, the most probable well-formed output in the target
language. A large language model asked to explain a passage will produce a
fluent, confident, complete gloss. Neither system has a representation for
“understood without being understood.” Neither can output a word and mark it as
carried but unpossessed. Fluency is the objective function; incompleteness is,
by construction, an error to be minimised.
Three specific losses follow when Shabnam meets
such a system, and each can be demonstrated to students in minutes.
First, script-flattening: The novel’s Arabic, Persian and
English arrive in Bengali script, which is the whole point: the reader sees a
Bengali page hosting foreign sound. Machine translation into English resolves
this. Fazl becomes “grace,” tamāmshud becomes “the end,” and the visible fact
of linguistic hospitality disappears into a monolingual surface. The two-word
structure that organises the ending is rendered invisible because both words
were English to begin with and the translation cannot show that they arrived as
guests.
Second, the closing of deferral: Asked what fazl and
rahmat mean, a language model will answer immediately and correctly. It cannot
easily produce the answer the novel gives, which is that the narrator has been
living with these words since childhood and still does not know. Summarisation
systems in particular are optimised to resolve; a passage whose content is an
unresolved question tends to be returned as a resolved statement.
Third, and most consequentially, asymmetry of resource:
Pratik Joshi and colleagues have documented the stark stratification of the
world’s languages within natural language processing, with a small set of
high-resource languages absorbing the overwhelming share of research attention
and data. Bengali, despite its enormous speaker base, is not among the best
served, and literary Bengali of the mid-twentieth century — with its
sādhu–calit alternation and dense Perso-Arabic lexis — is scarcer still. Emily
Bender and colleagues’ analysis of scale and training data in large language
models bears directly on what this means: a system’s apparent competence in a
language is a function of what it has seen, and the confident English sentence
it produces about a Bengali novel carries no signal about the thinness of the
corpus behind it.
This is where the critical AI literature earns its place
in a literary argument. Shakir Mohamed, Marie-Therese Png and William Isaac’s
framing of decolonial artificial intelligence identifies the structural risk:
systems built on metropolitan data distributions extend metropolitan epistemic
authority under the appearance of neutrality. Miranda Fricker’s concept of
hermeneutical injustice — the condition in which a community lacks the
interpretive resources to render its own experience intelligible — names the
effect at the receiving end. When a Bangladeshi undergraduate encounters Shabnam
through an English machine gloss, the gloss is not a neutral aid. It is an
interpretation produced by a system that has read a great deal of English
criticism and very little Bengali, and that is incapable of signalling the
difference.
The point is not that the machine is useless. It is that Shabnam
is an unusually sensitive instrument for detecting what the machine does. A
novel whose thesis is the productivity of incomplete understanding will
register, with unusual clarity, the operation of a technology built to
eliminate incompleteness.
VI. Teaching the Failure
The pedagogical conclusion is not prohibition. Students
in Bangladeshi English departments will use these tools; a rule against them is
unenforceable and, more importantly, wastes an opportunity. Because Shabnam
breaks the machine in an instructive way, the breakage can be made the object
of study. What follows is a sequence I have found defensible on both literary
and technological grounds, pitched at an undergraduate class reading the novel
in Bengali and writing about it in English.
The sequence begins with the untouched text. Students
read the fazl–rahmat passage in Bengali and write two hundred words, without
any digital assistance, on what the narrator claims to know and what he claims
not to know. This establishes a baseline that is theirs. They then submit the
same passage to a machine translation system and to a generative model, and
place the three accounts side by side — their own, the translation, the model’s
explanation — in a simple table. The analytical task is not to score the
machine but to catalogue difference: which words did it resolve that the novel
leaves open, which script-crossings vanished, what confidence it projected.
From there the class moves to the closing Bedouin tale
and repeats the exercise with a specific question: can any system be prompted
to preserve the fact that the girl does not understand the words she is
carrying? Students generally discover that it can be told to say so and cannot
be made to enact it — the output remains fluent English throughout, and the
asymmetry between the girl’s two words and the narrator’s learned Bengali
flattens out. That discovery is the lesson. It is also, usefully, a
transferable one: the same test can be run on any text whose meaning depends on
what it withholds.
The unit closes by returning to Apter’s and Venuti’s
terms and asking students to argue a position on whether tamāmshud should be
translated at all in a hypothetical English edition, and what a footnote costs.
David Damrosch’s account of world literature as a mode of circulation and
reading, rather than a canon, gives them a frame in which the question is
genuinely open rather than rhetorical. Assessment rests on the comparative
table and the final argument, both of which are difficult to produce without
having done the reading, and neither of which a model can complete on a
student’s behalf, since both require the student’s own prior, unassisted
encounter with the Bengali page.
There is a further, less technical gain. A generation of
students in Bangladesh is being taught, implicitly, that not understanding
something is a deficiency to be corrected by a query. Shabnam says
otherwise. It says that a woman who knows two words of a language may cross the
world on them, and that a man who knows twelve languages may fail to say the
one thing that matters. Teaching the novel well means teaching that
proposition, and the machine, by being unable to represent it, makes it
visible.
VII. Conclusion
Shabnam has
been read, when it has been read critically at all, as a romance of Mujtaba
Ali’s Afghan years and a companion piece to Deshe Bideshe. This article
has proposed a different account: that the novel is structured around a theory
of meaning as carried rather than possessed, that this theory is articulated
most sharply in its two paired episodes of two-word transmission, and that the
theory places the novel in direct and productive conflict with the assumptions
of contemporary language technology. That conflict is not a reason to keep the
technology out of the classroom. It is the reason to bring the novel in.
The wider implication concerns which texts will survive
the computational turn in a usable state. Works that yield cleanly to summary
will be increasingly encountered as summary. Works like Shabnam —
multilingual, allusive, organised around what they decline to explain — resist
that fate, but only if someone is teaching the resistance. Mujtaba Ali, who
spent his life moving between languages and wrote a novel about the
impossibility of arriving, would have recognised the situation. He left
instructions, of a sort, on his last page, in a language he did not translate.
Works Cited
Ali, Syed Mujtaba. Shabnam.Saiyad
Mujtaba Ali Rachanabali, vol. 5, Mitra o Ghosh Publishers, pp. 107–270.
Apter, Emily.Against
World Literature: On the Politics of Untranslatability. Verso, 2013.
Asad, Talal.“The Concept
of Cultural Translation in British Social Anthropology.”Writing Culture: The
Poetics and Politics of Ethnography, edited by James Clifford and George E.
Marcus, U of California P, 1986, pp. 141–64.
Bender, Emily M., et al.
“On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” Proceedings
of the 2021 ACM Conference on Fairness, Accountability, and Transparency,
Association for Computing Machinery, 2021, pp. 610–23.
Bhabha, Homi K. The
Location of Culture. Routledge, 1994.
Damrosch, David. What
Is World Literature? Princeton UP, 2003.
Fricker, Miranda. Epistemic
Injustice: Power and the Ethics of Knowing. Oxford UP, 2007.
Joshi, Pratik, et al.
“The State and Fate of Linguistic Diversity and Inclusion in the NLP World.” Proceedings
of the 58th Annual Meeting of the Association for Computational Linguistics,
Association for Computational Linguistics, 2020, pp. 6282–93.
Mohamed, Shakir,
Marie-Therese Png, and William Isaac. “Decolonial AI: Decolonial Theory as
Sociotechnical Foresight in Artificial Intelligence.” Philosophy and Technology,
vol. 33, no. 4, 2020, pp. 659–84.
Pollock, Sheldon.
“Cosmopolitan and Vernacular in History.” Public Culture, vol. 12, no.
3, 2000, pp. 591–625.
Said, Edward W. Orientalism.
Pantheon Books, 1978.
Venuti, Lawrence. The
Translator’s Invisibility: A History of Translation. Routledge, 1995.