Yeah, that is a bit of a nasty question :-) because its well known that Oracle does not reveal the column name for number datatypes.

I suspect the intent of this interview question was to get you talking about how to tackle it...so some ideas would be:

1) If its a simple literal statement, you can use DBMS_SQL to get the column in error

SQL> create table t
  2   ( x number(2),
  3     y number(3),
  4     z number(4)
  5   );

Table created.

SQL>
SQL> insert into t values (123,123,123);
insert into t values (123,123,123)
                      *
ERROR at line 1:
ORA-01438: value larger than specified precision allowed for this column


SQL>
SQL> set serverout on
SQL> declare
  2    c   integer := dbms_sql.open_cursor ();
  3    erg integer;
  4    l_sql varchar2(4000) := 'insert into t values (123,123,123)';
  5  begin
  6    dbms_sql.parse (c, l_sql, dbms_sql.native);
  7    erg := dbms_sql.execute(c);
  8    dbms_sql.close_cursor (c);
  9  exception
 10    when others then
 11      dbms_output.put_line(sqlerrm);
 12      dbms_output.put_line(l_sql);
 13      dbms_output.put_line(lpad('^', dbms_sql.last_error_position+1,'='));
 14      dbms_sql.close_cursor (c);
 15  end;
 16  /
ORA-01438: value larger than specified precision allowed for this column
insert into t values (123,123,123)
======================^

PL/SQL procedure successfully completed.

2) In a more realistic scenarios, the values aren't known in advance. For that one, a SQL trace to capture binds can assist, eg

exec dbms_monitor.session_trace_enable(binds=>true)
[run sql]

The trace file contains each bind variable value and that can be compared to the datatype sizes for each column